feat(flags): put /about and /blog behind flags, dark by default

Both features ship dark. Neither is reachable in an environment where
its flag is off, and every flag in this system starts off, so a deploy
of this commit makes both disappear until someone turns them on
deliberately.

Two independent flags rather than one, which makes blog-on-about-off a
reachable state. That state is the whole reason the change is larger
than four notFound() calls: the blog leans on the About page for its
author identity. The Person entity is anchored at /about#tudor, and
that URL 404s while about_page is dark, so a post published in that
state would claim an author resolving to nothing. Worse than having no
named author. Both bylines fall back to unlinked text and the
BlogPosting attributes to the publisher instead, so every combination
of the two flags renders something correct.

Gated: /about, /blog, /blog/[slug], the RSS feed, both footer links,
and the matching content-sitemap entries. A sitemap must never
advertise a URL that 404s. With both dark it emits a valid empty
urlset rather than a 404, because robots.txt names it unconditionally.

Not gated: /admin. Posts have to be writable before the blog is worth
switching on, so flagging the panel would make the flag unflippable.

getFlags takes a revalidate rather than always using the 300s
constant. Reading a flag pins the calling route to the lowest
revalidate among its fetches, and the footer links live in the root
layout, so a naive gate there would have dropped every school and
place page from a weekly cache to a 5-minute one. The layout passes
604800, the floor those routes already declare, and the build confirms
all four SSG route families still prerender. The cost is one-way
latency: pages follow a flip in minutes, footer links within a week.

The e2e journeys follow the existing paired shape from the
admission_distance flag: a lit journey and a dark one for each flag,
reading state from whether /about and /blog respond rather than from
/api/flags, which another journey asserts is not publicly reachable.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DXnXQKnPpZBBP61fBQiFkq

backend/app.py

@@ -6,6 +6,7 @@ Uses real data from UK Government Compare School Performance downloads.
 
 import hashlib
 import re
+import time
 from contextlib import asynccontextmanager
 from datetime import datetime, timezone
 from typing import Optional
@@ -15,7 +16,7 @@ import pandas as pd
 from fastapi import FastAPI, HTTPException, Query, Request, Depends, Header
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.middleware.gzip import GZipMiddleware
-from fastapi.responses import FileResponse, Response
+from fastapi.responses import FileResponse, JSONResponse, Response
 from fastapi.staticfiles import StaticFiles
 from slowapi import Limiter, _rate_limit_exceeded_handler
 from slowapi.util import get_remote_address
@@ -33,8 +34,10 @@ from .data_loader import (
     get_supplementary_data,
     get_supplementary_data_batch,
     search_schools_typesense,
+    suggest_schools_typesense,
 )
 from .data_loader import get_data_info as get_db_info
+from . import flags
 from .places import build_place_registry
 from .schemas import METRIC_DEFINITIONS, RANKING_COLUMNS, SCHOOL_COLUMNS
 from .utils import clean_for_json, convert_to_native
@@ -222,10 +225,10 @@ def _place_sitemap_rows(kinds: tuple[str, ...]) -> list[str]:
         if p.kind not in kinds:
             continue
         rows.append(_url_element(BASE_URL + _place_url(p)))
-        # Outcodes carry no phase variants: nobody searches "primary schools
-        # in SW11", so the routes do not exist to submit.
-        if p.kind == "outcode":
-            continue
+        # Which phases a place publishes is the registry's decision alone —
+        # outcodes report none, because the spec gives them no phase route.
+        # Repeating that rule here was how the page and the sitemap came to
+        # disagree about which URLs exist.
         for phase in ("primary", "secondary"):
             if p.publishes_phase(phase):
                 rows.append(_url_element(f"{BASE_URL}{_place_url(p)}/{phase}"))
@@ -295,8 +298,101 @@ def clean_filter_values(series: pd.Series) -> list[str]:
 # SECURITY MIDDLEWARE & HELPERS
 # =============================================================================
 
-# Rate limiter
-limiter = Limiter(key_func=get_remote_address)
+def client_key(request: Request) -> str:
+    """The rate-limit bucket: the real caller, not the proxy in front of them.
+
+    `get_remote_address` reads request.client.host. In staging and production
+    the backend has no published ports and sits on the internal network, so its
+    only caller is the Next proxy — meaning every browser user on the site
+    shared one bucket. Measured before this fix: 70 concurrent requests to
+    /api/schools returned 60 OK and 10 refused.
+
+    CF-Connecting-IP first, because Cloudflare (in front of both environments)
+    sets it on every origin request and *overwrites* any client-supplied value,
+    which a parsed X-Forwarded-For chain does not guarantee. The XFF fallback is
+    forgeable, but only by a caller already inside the Docker network, which is
+    the one place nothing untrusted can reach.
+    """
+    cf = request.headers.get("cf-connecting-ip")
+    if cf:
+        return cf.strip()
+    xff = request.headers.get("x-forwarded-for")
+    if xff:
+        return xff.split(",")[0].strip()
+    return get_remote_address(request)
+
+
+# Per-client limiter. Paired with the global ceiling below — the two do
+# different jobs and neither substitutes for the other.
+limiter = Limiter(key_func=client_key)
+
+
+# --- The ceiling no header can raise ----------------------------------------
+#
+# client_key trusts CF-Connecting-IP, and nothing in this process can tell an
+# edge-set header from an attacker-set one. That distinction can only be made
+# at Cloudflare, with Authenticated Origin Pulls or an origin firewall. A
+# caller reaching the origin directly could otherwise mint a fresh rate-limit
+# bucket per request and evade per-client limits entirely — which would make
+# correct keying a net regression against abuse, since the single shared bucket
+# it replaced at least capped everyone at 60/minute together.
+#
+# So per-client limits give fairness, and this gives the origin a hard total.
+# It does not make the header trustworthy; it bounds what trusting it can cost.
+# The header problem itself is closed at Cloudflare, not here.
+#
+# [window_start_monotonic, count], or None before the first request. A fixed
+# window is crude, which is right for a backstop: it has to be obviously
+# correct rather than fair.
+_global_window: Optional[list] = None
+
+# The container healthcheck runs `curl http://localhost:80/api/data-info` from
+# inside the container. Starving it would fail the check, restart the
+# container, and turn a load spike into an outage loop — the ceiling exists to
+# protect the origin, not to kill it.
+_LOCAL_HOSTS = frozenset({"127.0.0.1", "::1", "localhost"})
+
+
+def exempt_from_ceiling(request: Request) -> bool:
+    """Whether the ceiling should ignore this request.
+
+    Its own function so the rule is testable without standing up a server —
+    and so the healthcheck exemption is somewhere a reader can find it.
+    """
+    if not request.url.path.startswith("/api/"):
+        return True
+    # The peer address, never the Host header: Host is set by the caller and
+    # would hand every attacker an exemption.
+    return (request.client.host if request.client else "") in _LOCAL_HOSTS
+
+
+class GlobalRateLimitMiddleware(BaseHTTPMiddleware):
+    """A cap on total /api/ traffic, independent of any client identity."""
+
+    async def dispatch(self, request: Request, call_next):
+        global _global_window
+
+        if exempt_from_ceiling(request):
+            return await call_next(request)
+
+        now = time.monotonic()
+        # One event loop, and no await between the read and the write, so this
+        # sequence is atomic without a lock.
+        if _global_window is None or now - _global_window[0] >= 60:
+            _global_window = [now, 0]
+        _global_window[1] += 1
+
+        if _global_window[1] > settings.global_rate_limit_per_minute:
+            return JSONResponse(
+                # Distinguishable from slowapi's per-client 429: an operator
+                # reading logs has to be able to tell "one noisy client" from
+                # "the origin is saturated".
+                {"detail": "The service is at capacity. Please retry shortly."},
+                status_code=429,
+                headers={"Retry-After":
+                         str(max(1, int(60 - (now - _global_window[0]))))},
+            )
+        return await call_next(request)
 
 
 class SecurityHeadersMiddleware(BaseHTTPMiddleware):
@@ -354,6 +450,7 @@ CACHE_RULES: list[tuple[str, tuple[int, int, int]]] = [
     ("/api/schools/",           (300,  3600,  86400)),  # /api/schools/{urn}
     ("/api/rankings",           (60,   600,   3600)),
     ("/api/compare",            (60,   600,   3600)),
+    ("/api/suggest",            (60,   3600,  86400)),  # autosuggest
     ("/api/schools",            (30,   300,   1800)),   # search list
 ]
 
@@ -459,6 +556,7 @@ def validate_postcode(postcode: Optional[str]) -> Optional[str]:
 async def lifespan(app: FastAPI):
     """Application lifespan - startup and shutdown events."""
     global _sitemaps
+    flags.init()
     print("Loading school data from marts...")
     df = load_school_data()
     if df.empty:
@@ -500,6 +598,10 @@ app.add_middleware(CacheAndETagMiddleware)
 app.add_middleware(SecurityHeadersMiddleware)
 app.add_middleware(RequestSizeLimitMiddleware)
 app.add_middleware(GZipMiddleware, minimum_size=512)
+# Added last, so it is outermost and refuses before anything downstream does
+# work. A ceiling that only applies after the expensive part has run is not a
+# ceiling.
+app.add_middleware(GlobalRateLimitMiddleware)
 
 # CORS middleware - restricted for production
 app.add_middleware(
@@ -806,11 +908,18 @@ async def get_school_details(request: Request, urn: int):
         "census": supplementary.get("census"),
         "admissions": supplementary.get("admissions"),
         "admissions_history": supplementary.get("admissions_history") or [],
-        "admission_distance": supplementary.get("admission_distance"),
+        # Behind a flag, and withheld at the source rather than rendered-but-
+        # hidden: this endpoint is public and unauthenticated, so a field left
+        # in the payload is a published field. The key is absent, not null —
+        # null would state that this school has no cut-off, which is a
+        # different claim from "we are not publishing cut-offs".
+        **({"admission_distance": supplementary.get("admission_distance")}
+           if flags.is_enabled("admission_distance") else {}),
         "sen_detail": supplementary.get("sen_detail"),
         "phonics": supplementary.get("phonics"),
         "deprivation": supplementary.get("deprivation"),
         "finance": supplementary.get("finance"),
+        "destinations": supplementary.get("destinations"),
     }
 
 
@@ -1200,7 +1309,8 @@ async def get_place(request: Request, kind: str, slug: str,
     if kind not in VALID_PLACE_KINDS:
         raise HTTPException(status_code=404, detail="No such place")
 
-    place = get_place_registry().get(f"{kind}:{slug}")
+    registry = get_place_registry()
+    place = registry.get(f"{kind}:{slug}")
     if place is None:
         raise HTTPException(status_code=404, detail="No such place")
 
@@ -1212,10 +1322,15 @@ async def get_place(request: Request, kind: str, slug: str,
         if wanted and "phase" in rows.columns:
             rows = rows[rows["phase"].fillna("").str.lower().isin(wanted)]
 
-    # The metric the page ranks on, which is also the one it averages.
+    # The metric the page shows, and averages.
     metric = "attainment_8_score" if phase == "secondary" else "rwm_expected_pct"
-    if metric in rows.columns:
-        rows = rows.sort_values(metric, ascending=False, na_position="last")
+
+    # Alphabetical, not by score. A place page is read by someone looking for
+    # a school they can name, and scanning for it is what the order should
+    # serve. /rankings is where the league-table ordering lives, and it keeps
+    # sorting by metric.
+    if "school_name" in rows.columns:
+        rows = rows.sort_values("school_name", key=lambda c: c.str.lower())
 
     averages = {
         m: (None if m not in rows.columns or rows[m].dropna().empty
@@ -1223,9 +1338,22 @@ async def get_place(request: Request, kind: str, slug: str,
         for m in ("rwm_expected_pct", "attainment_8_score")
     }
 
-    cols = [c for c in SCHOOL_COLUMNS + ["latitude", "longitude", "phase",
-                                          "rwm_expected_pct", "attainment_8_score",
-                                          "total_pupils"]
+    # dict.fromkeys, not a list: SCHOOL_COLUMNS already ends with latitude and
+    # longitude, so concatenating them again selected each twice and pandas
+    # dropped one of every duplicated pair with a "columns are not unique"
+    # warning. Ordered de-duplication keeps the column order and the warning
+    # cannot come back.
+    #
+    # nursery_provision and parliamentary_constituency are not in
+    # SCHOOL_COLUMNS and the place table shows both. The `in rows.columns`
+    # guard is what keeps a mart the pipeline has not rebuilt working: those
+    # two are the optional GIAS columns data_loader degrades to NULL.
+    cols = [c for c in dict.fromkeys(
+                SCHOOL_COLUMNS + ["latitude", "longitude", "phase",
+                                  "nursery_provision",
+                                  "parliamentary_constituency",
+                                  "rwm_expected_pct", "attainment_8_score",
+                                  "total_pupils"])
             if c in rows.columns]
 
     return {
@@ -1235,8 +1363,17 @@ async def get_place(request: Request, kind: str, slug: str,
                   # Every authority the place meaningfully sits in. SW19 is
                   # mostly Merton but partly Wandsworth; naming one asserts
                   # something false.
+                  #
+                  # The slug is null where that authority has no page of its
+                  # own: City of London and the Isles of Scilly hold fewer
+                  # schools than the threshold. Naming them is still right;
+                  # linking them would be a 404.
                   "authorities": [
-                      {"name": name, "slug": _slugify(name), "count": n}
+                      {"name": name,
+                       "slug": (_slugify(name)
+                                if f"authority:{_slugify(name)}" in registry
+                                else None),
+                       "count": n}
                       for name, n in place.authorities
                   ],
                   # Only phases that clear the threshold, so the page links
@@ -1248,6 +1385,50 @@ async def get_place(request: Request, kind: str, slug: str,
     }
 
 
+# Two characters. One is not a query — it matches thousands of schools and the
+# response is useless, so it is not worth a round trip.
+SUGGEST_MIN_QUERY = 2
+
+
+@app.get("/api/suggest")
+@limiter.limit("120/minute")
+async def suggest_schools(
+    request: Request,
+    q: str = Query("", max_length=100),
+    limit: int = Query(8, ge=1, le=20),
+):
+    """School name suggestions, from Typesense alone.
+
+    Deliberately not a mode of /api/schools: that path filters and sorts the
+    full in-memory DataFrame, which is far too expensive to run per keystroke.
+
+    Nothing here returns an error for ordinary input. A short query, no
+    matches, or Typesense being unreachable are all 200 with an empty list —
+    a dropdown that quietly does not appear is the right failure for a
+    keystroke path, and there is no DataFrame fallback because the 25,000-row
+    substring scan is precisely what this endpoint exists to avoid.
+
+    120/minute rather than the default 60: a 200 ms debounce makes typing
+    legitimately bursty.
+    """
+    query = q.strip()
+    if len(query) < SUGGEST_MIN_QUERY:
+        return {"suggestions": []}
+    return {"suggestions": suggest_schools_typesense(query, limit)}
+
+
+@app.get("/api/flags")
+@limiter.limit(f"{settings.rate_limit_per_minute}/minute")
+async def get_feature_flags(request: Request):
+    """Every declared flag and its current value.
+
+    Internal only. The Next proxy denies this path, because the response names
+    every unreleased feature the codebase knows about — which is exactly what
+    shipping dark is meant to keep quiet.
+    """
+    return flags.all_flags()
+
+
 @app.get("/api/data-info")
 @limiter.limit(f"{settings.rate_limit_per_minute}/minute")
 async def get_data_info(request: Request):

backend/config.py

@@ -35,6 +35,11 @@ class Settings(BaseSettings):
     # Security
     admin_api_key: str = Field(default_factory=lambda: secrets.token_urlsafe(32))
     rate_limit_per_minute: int = 60  # Requests per minute per IP
+    # A ceiling on total /api/ traffic, independent of any client identity.
+    # client_key trusts headers only Cloudflare can vouch for, so a caller
+    # reaching the origin directly could otherwise mint a fresh bucket per
+    # request. See GlobalRateLimitMiddleware in backend/app.py.
+    global_rate_limit_per_minute: int = 3000
     rate_limit_burst: int = 10  # Allow burst of requests
     max_request_size: int = 1024 * 1024  # 1MB max request size
 
@@ -42,6 +47,16 @@ class Settings(BaseSettings):
     typesense_url: str = "http://localhost:8108"
     typesense_api_key: str = ""
 
+    # Feature flags (Unleash). An empty unleash_url disables flags entirely and
+    # every flag evaluates False — the correct behaviour for local development
+    # and CI, and the reason no test needs a running Unleash.
+    unleash_url: str = ""
+    unleash_api_token: str = ""
+    unleash_app_name: str = "schoolcompare-backend"
+    # On a named volume, so a restart during an Unleash outage keeps
+    # last-known state instead of reverting a released feature to dark.
+    unleash_cache_directory: str = "/app/.unleash"
+
     # Analytics
     ga_measurement_id: Optional[str] = "G-J0PCVT14NY"  # Google Analytics 4 Measurement ID
 

backend/data_loader.py

@@ -20,6 +20,7 @@ from .models import (
     DimSchool, DimLocation, KS2Performance,
     FactOfstedInspection, FactAdmissions, FactAdmissionDistance,
     FactDeprivation, FactFinance, FactPupilCharacteristics,
+    FactKs4Destinations, FactKs5Destinations,
 )
 from .ofsted_codes import ofsted_page_url, report_card_labels
 from .schemas import SCHOOL_TYPE_MAP
@@ -100,6 +101,58 @@ def search_schools_typesense(query: str, limit: int = 250) -> List[int]:
         return []
 
 
+# The most a public endpoint will return in one response.
+SUGGEST_MAX_LIMIT = 20
+
+# Fields a suggestion row carries, and the default when the document omits an
+# optional one. phase and school_type are optional in the Typesense schema.
+_SUGGEST_FIELDS = ("school_name", "local_authority", "postcode",
+                   "phase", "school_type")
+
+
+def suggest_schools_typesense(query: str, limit: int = 8) -> List[dict]:
+    """Autosuggest rows straight from Typesense. Never raises.
+
+    Returns documents rather than URNs, unlike search_schools_typesense, so the
+    caller needs no DataFrame. Every field below is already in the index — see
+    pipeline/scripts/sync_typesense.py — which is what makes this cheap enough
+    to run per keystroke.
+    """
+    client = _get_typesense_client()
+    if client is None:
+        return []
+    try:
+        result = client.collections["schools"].documents.search({
+            "q": query,
+            "query_by": "school_name,local_authority",
+            "per_page": max(1, min(limit, SUGGEST_MAX_LIMIT)),
+            "typo_tokens_threshold": 1,
+        })
+    except Exception:
+        # A dropdown that quietly stops appearing is the right failure here.
+        return []
+
+    rows = []
+    for hit in result.get("hits", []) or []:
+        doc = (hit or {}).get("document") or {}
+        try:
+            urn = int(doc["urn"])
+        except (KeyError, TypeError, ValueError):
+            # Skip the row, keep the rest. Typesense declares urn as int32 so
+            # this should be unreachable, but the index is a separate system
+            # that something other than this code can reindex — and "never
+            # raises" is a promise the keystroke path actually depends on.
+            # Dropping one malformed document is right; blanking the whole
+            # dropdown, or serving a suggestion pointing at /school/0, is not.
+            logging.getLogger(__name__).warning(
+                "skipping malformed suggestion document: %r", doc)
+            continue
+        row = {"urn": urn}
+        row.update({f: str(doc.get(f, "") or "") for f in _SUGGEST_FIELDS})
+        rows.append(row)
+    return rows
+
+
 def normalize_school_type(school_type: Optional[str]) -> Optional[str]:
     """Convert cryptic school type codes to user-friendly names."""
     if not school_type:
@@ -764,6 +817,218 @@ def _finance_dict(f) -> dict:
     }
 
 
+# Destination measures that are totals DfE published itself, rather than one of
+# the categories that partition the cohort.
+_AGGREGATE_MEASURES = {"agg_sustained_education", "agg_sustained_all"}
+
+
+def _format_cohort_year(year) -> str | None:
+    """202223 -> '2022/23'.
+
+    The section has to date its own cohort. Destination measures run about two
+    GCSE years behind the results shown above them on the same page, so an
+    undated figure reads as stale data rather than as a different question.
+    """
+    if not year:
+        return None
+    text = str(year)
+    if len(text) == 6:
+        return f"{text[:4]}/{text[4:6]}"
+    if len(text) == 8:
+        return f"{text[:4]}/{text[6:8]}"
+    return text
+
+
+_PUPIL_GROUPS = ("disadvantaged", "other", "all")
+
+
+def _lone_hidden_groups(groups: dict) -> list:
+    """Pupil groups hiding exactly one category — solvable by subtraction."""
+    return [
+        key for key, group in groups.items()
+        if sum(1 for c in group["categories"] if c["status"] == "suppressed") == 1
+    ]
+
+
+def _lone_hidden_categories(groups: dict) -> list:
+    """Categories hidden in exactly one of several pupil groups."""
+    lone = []
+    categories = {c["category"] for g in groups.values() for c in g["categories"]}
+    for category in categories:
+        found = [
+            c for g in groups.values() for c in g["categories"]
+            if c["category"] == category
+        ]
+        hidden = [c for c in found if c["status"] == "suppressed"]
+        if len(hidden) == 1 and len(found) > 1:
+            lone.append(category)
+    return lone
+
+
+def disclosure_invariant_holds(groups: dict) -> bool:
+    """Every row and every column hides none, or at least two.
+
+    Public so the tests can assert it directly rather than re-deriving it.
+    """
+    return not _lone_hidden_groups(groups) and not _lone_hidden_categories(groups)
+
+
+def _mask_for_disclosure(groups: dict) -> None:
+    """Withhold further cells until nothing suppressed can be solved for.
+
+    Not rendering a figure is not the same as not publishing it. This endpoint
+    is public and unauthenticated, so anything left in the payload is
+    published, whatever the UI chooses to draw — the same reasoning the
+    admission_distance field carries in app.py.
+
+    Two identities let a caller solve for a withheld cell:
+
+      * within a pupil group, the categories sum to the cohort, so a group with
+        exactly ONE suppressed category gives it away as cohort - sum(rest);
+      * across groups, disadvantaged + other = all for every category, so a
+        category suppressed in exactly ONE of the three gives itself away.
+
+    DfE's own answer is secondary suppression: withhold a second cell so the
+    residual spans two unknowns and identifies neither.
+
+    Where no companion can do that — a sparse cohort whose every other category
+    is `not_applicable`, which is common in special schools and alternative
+    provision — there is nothing left to withhold, so the pupil group is
+    DROPPED entirely. An earlier version simply gave up here and returned with
+    the violation intact and no signal, which is the one outcome this function
+    must never produce: a disclosure-control pass that fails silently is worse
+    than none, because everything downstream trusts it.
+
+    Mutates `groups` in place. Guaranteed to return with
+    disclosure_invariant_holds(groups) true.
+    """
+
+    def suppress(cell):
+        if cell["status"] == "published":
+            cell["status"] = "suppressed"
+            cell["pupils"] = None
+            cell["percentage"] = None
+            return True
+        return False
+
+    def add_companion(candidates) -> bool:
+        """Withhold a second cell so the residual spans two unknowns.
+
+        The companion must carry pupils. Suppressing a zero looks like
+        secondary suppression and protects nothing: the residual still equals
+        the original withheld figure exactly. Returns False when no cell can
+        do the job, which escalates to dropping the group.
+        """
+        published = [c for c in candidates if c["status"] == "published"]
+        useful = sorted(
+            (c for c in published if (c["pupils"] or 0) > 0),
+            key=lambda c: c["pupils"],
+        )
+        if useful:
+            return suppress(useful[0])
+        # Every remaining cell is zero or not applicable: withholding any of
+        # them leaves the residual equal to the original figure.
+        return False
+
+    # Fixpoint: each new suppression can break the other identity. Terminates
+    # because every pass either adds a suppression, drops a group, or stops.
+    while not disclosure_invariant_holds(groups):
+        changed = False
+
+        for category in _lone_hidden_categories(groups):
+            siblings = [
+                c for g in groups.values() for c in g["categories"]
+                if c["category"] == category
+            ]
+            if add_companion(siblings):
+                changed = True
+
+        for key in _lone_hidden_groups(groups):
+            if add_companion(groups[key]["categories"]):
+                changed = True
+
+        if changed:
+            continue
+
+        # Nothing left to withhold. Drop the groups that are still solvable,
+        # and any category still solvable across the groups that remain.
+        for key in _lone_hidden_groups(groups):
+            del groups[key]
+            changed = True
+
+        for category in _lone_hidden_categories(groups):
+            for group in groups.values():
+                for cell in group["categories"]:
+                    if cell["category"] == category and suppress(cell):
+                        changed = True
+
+        if not changed:
+            # Unreachable given the two escalations above, but a masking pass
+            # must never spin or exit unsafely. Withhold everything.
+            groups.clear()
+            return
+
+
+def _destinations_block(rows: list) -> dict | None:
+    """Shape destination rows for one phase into the API's block.
+
+    Applies secondary suppression before returning, so no caller of this public
+    endpoint can solve for a figure DfE withheld. See _mask_for_disclosure.
+
+    Aggregate measures are dropped entirely. DfE publishes them, and they would
+    be useful for a "what is published for this group" fallback, but nothing
+    renders them today and an aggregate spanning exactly one suppressed
+    component names that component. An unused field that leaks is not a
+    trade-off worth carrying — re-add them with their own guard if the fallback
+    is ever built.
+
+    Deliberately computes no residual, no "remaining pupils" figure, and no
+    total that would close a gap left by a suppressed category.
+    """
+    if not rows:
+        return None
+
+    years = [r["year"] for r in rows if r.get("year") is not None]
+    if not years:
+        return None
+    latest_year = max(years)
+    rows = [r for r in rows if r.get("year") == latest_year]
+
+    groups: dict = {}
+    for row in rows:
+        group = groups.setdefault(
+            row["pupil_group"],
+            {"cohort": row.get("cohort_pupils"), "categories": []},
+        )
+        measure = row["destination_measure"]
+        published = row.get("status") == "published"
+        # Belt and braces: percentage is derived from the same source cell as
+        # pupils, but publishing one without the other would hand back the
+        # cohort (pupils / percentage) and with it the residual.
+        cell = {
+            "category": measure,
+            "pupils": row.get("pupils") if published else None,
+            "percentage": row.get("percentage") if published else None,
+            "status": row.get("status"),
+        }
+        if measure in _AGGREGATE_MEASURES:
+            continue
+        group["categories"].append(cell)
+
+    if not groups:
+        return None
+
+    _mask_for_disclosure(groups)
+
+    # Masking can empty the block entirely — a sparse cohort where no group
+    # could be made safe. Return None so the section is absent rather than
+    # rendering an empty shell.
+    if not groups:
+        return None
+
+    return {"cohort_year": _format_cohort_year(latest_year), "groups": groups}
+
+
 def _empty_supplementary() -> dict:
     return {
         "ofsted": None,
@@ -775,6 +1040,7 @@ def _empty_supplementary() -> dict:
         "phonics": None,
         "deprivation": None,
         "finance": None,
+        "destinations": None,
     }
 
 
@@ -902,6 +1168,38 @@ def get_supplementary_data_batch(db: Session, urns: list[int]) -> dict:
             result[f.urn]["finance"] = _finance_dict(f)
     _safe(_finance)
 
+    # Destinations — KS4 and 16-18. Both marts are long-format, so every row
+    # for a URN is collected and _destinations_block picks the latest year and
+    # shapes the pupil groups. A phase with no rows serialises as null rather
+    # than an empty shell, so the frontend renders nothing rather than an empty
+    # section.
+    def _destinations():
+        from collections import defaultdict
+
+        def _collect(model):
+            per_urn = defaultdict(list)
+            for r in db.query(model).filter(model.urn.in_(urns)).all():
+                per_urn[r.urn].append({
+                    "year": r.year,
+                    "pupil_group": r.pupil_group,
+                    "destination_measure": r.destination_measure,
+                    "cohort_pupils": r.cohort_pupils,
+                    "pupils": r.pupils,
+                    "percentage": r.percentage,
+                    "status": r.status,
+                })
+            return per_urn
+
+        ks4_rows = _collect(FactKs4Destinations)
+        ks5_rows = _collect(FactKs5Destinations)
+        for urn in urns:
+            ks4 = _destinations_block(ks4_rows.get(urn, []))
+            ks5 = _destinations_block(ks5_rows.get(urn, []))
+            result[urn]["destinations"] = (
+                {"ks4": ks4, "ks5": ks5} if (ks4 or ks5) else None
+            )
+    _safe(_destinations)
+
     return result
 
 

backend/flags.py

@@ -0,0 +1,135 @@
+"""Feature flags: what can be switched, and what is switched right now.
+
+Ship-dark, not a kill switch. Flags let work merge and deploy without becoming
+visible; they are expected to flip about monthly, by a person, deliberately.
+Nothing here does percentage rollouts or user targeting — the site has no user
+identity to target.
+
+Unleash holds the state. It does not hold the list. REGISTRY below is that
+list, and it exists for three reasons: the SDK evaluates an unknown flag to
+False, so without a registry that is an *undeclared* False, indistinguishable
+from a typo; /api/flags needs a key set to return when Unleash is unreachable;
+and a flag in the UI but not in the registry is orphaned and should be visibly
+so rather than quietly authoritative.
+
+Every flag defaults to False. There is no per-flag default, because a flag that
+defaults on is a kill switch, and this is not one.
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass
+from datetime import date
+
+from .config import settings
+
+logger = logging.getLogger(__name__)
+
+# A flag is temporary scaffolding. See test_a_flag_older_than_the_limit.
+MAX_FLAG_AGE_DAYS = 90
+
+
+@dataclass(frozen=True)
+class Flag:
+    # One string: the registry key, the Unleash flag name, and the JSON key in
+    # /api/flags. snake_case, matching the API's existing convention. No case
+    # transformation anywhere, so there is no mapping layer to get wrong.
+    name: str
+    description: str          # one line: what turning this on reveals
+    added: date               # for the staleness tripwire
+
+
+REGISTRY: dict[str, Flag] = {
+    f.name: f for f in (
+        Flag(
+            name="admission_distance",
+            description=(
+                "The last-distance-offered figure on the Admissions tile and "
+                "the 'How far away are you?' section on school pages."
+            ),
+            added=date(2026, 8, 23),
+        ),
+        Flag(
+            name="school_autosuggest",
+            description=(
+                "School name suggestions as you type in the main search box."
+            ),
+            added=date(2026, 8, 26),
+        ),
+        Flag(
+            name="about_page",
+            description=(
+                "The /about page, its footer link, its sitemap entry, and the "
+                "named-author byline on every blog post."
+            ),
+            added=date(2026, 9, 8),
+        ),
+        Flag(
+            name="blog",
+            description=(
+                "The /blog index, post pages, the RSS feed, their footer link "
+                "and their sitemap entries. Not /admin: posts must be "
+                "writable before the blog is readable."
+            ),
+            added=date(2026, 9, 8),
+        ),
+    )
+}
+
+
+_client = None
+
+
+def init() -> None:
+    """Start the Unleash client, or log why flags are all off.
+
+    Called once from the app lifespan. Never raises: a flag system that can
+    stop the API from booting is worse than one that is switched off.
+    """
+    global _client
+    if not settings.unleash_url or not settings.unleash_api_token:
+        logger.warning(
+            "Unleash is not configured (UNLEASH_URL / UNLEASH_API_TOKEN); "
+            "every feature flag evaluates to False.")
+        return
+
+    try:
+        from UnleashClient import UnleashClient
+
+        _client = UnleashClient(
+            url=settings.unleash_url,
+            app_name=settings.unleash_app_name,
+            custom_headers={"Authorization": settings.unleash_api_token},
+            cache_directory=settings.unleash_cache_directory,
+            refresh_interval=15,
+        )
+        _client.initialize_client()
+        logger.info("Unleash client initialised against %s", settings.unleash_url)
+    except Exception:
+        # Fail closed and keep serving. The SDK also evaluates everything False
+        # until its first successful sync, so this is the same direction.
+        _client = None
+        logger.exception("Unleash client failed to start; flags are all False.")
+
+
+def is_enabled(name: str) -> bool:
+    """Whether `name` is on. False for anything unknown, unreachable or broken."""
+    if name not in REGISTRY:
+        logger.error(
+            "undeclared feature flag %r was evaluated; returning False. "
+            "Add it to backend/flags.py REGISTRY or fix the name.", name)
+        return False
+    if _client is None:
+        return False
+    try:
+        return bool(_client.is_enabled(
+            name, fallback_function=lambda feature_name, context: False))
+    except Exception:
+        logger.exception("flag %r failed to evaluate; returning False", name)
+        return False
+
+
+def all_flags() -> dict[str, bool]:
+    """Every declared flag and its current value. Serves /api/flags."""
+    return {name: is_enabled(name) for name in REGISTRY}

backend/models.py

@@ -321,3 +321,48 @@ class Ks2NationalAverage(Base):
     gps_high_pct = Column(Float)
     gps_avg_score = Column(Float)
     science_expected_pct = Column(Float)
+
+
+class FactKs4Destinations(Base):
+    """KS4 leavers destinations — one row per URN, year, pupil group, measure.
+
+    Long format rather than wide because pupil_group is a real third dimension.
+    `status` is load-bearing: 'suppressed' means DfE withheld a figure it
+    considered disclosive and the page must print "withheld"; 'not_applicable'
+    means the measure does not apply and the page must print nothing. `pupils`
+    is null for both, so collapsing status to a null check loses the
+    difference — and the categories sum to the cohort, so a consumer that
+    treats a withheld cell as zero republishes what DfE hid.
+    """
+    __tablename__ = "fact_ks4_destinations"
+    __table_args__ = (
+        Index("ix_ks4_dest_urn_year", "urn", "year"),
+        MARTS,
+    )
+
+    urn = Column(Integer, primary_key=True)
+    year = Column(Integer, primary_key=True)
+    pupil_group = Column(String(20), primary_key=True)
+    destination_measure = Column(String(40), primary_key=True)
+    cohort_pupils = Column(Integer)
+    pupils = Column(Integer)
+    percentage = Column(Float)
+    status = Column(String(20))
+
+
+class FactKs5Destinations(Base):
+    """16-18 study leavers destinations — same grain as FactKs4Destinations."""
+    __tablename__ = "fact_ks5_destinations"
+    __table_args__ = (
+        Index("ix_ks5_dest_urn_year", "urn", "year"),
+        MARTS,
+    )
+
+    urn = Column(Integer, primary_key=True)
+    year = Column(Integer, primary_key=True)
+    pupil_group = Column(String(20), primary_key=True)
+    destination_measure = Column(String(40), primary_key=True)
+    cohort_pupils = Column(Integer)
+    pupils = Column(Integer)
+    percentage = Column(Float)
+    status = Column(String(20))

backend/places.py

@@ -59,18 +59,51 @@ def _publishable_urns(df) -> set[int]:
     return set(df.loc[df[cols].notna().any(axis=1), "urn"].astype(int))
 
 
+# The measure a phase page is built around. A page with no results in this
+# column has nothing a list of school names does not already give.
+_PHASE_METRIC = {
+    "primary": "rwm_expected_pct",
+    "secondary": "attainment_8_score",
+}
+
+
 def _phase_urns(group, publishable: set[int]) -> dict[str, tuple[int, ...]]:
-    """URNs per phase. All-through schools count toward both, matching the
-    PHASE_GROUPS mapping the search filters already use."""
+    """URNs per phase, counting only schools with a result for that phase.
+
+    Not merely "publishable". A school with an Ofsted grade and no results is
+    worth a page of its own and belongs in the place list, but it cannot
+    populate a phase page's results column — and the threshold is there to ask
+    whether that column will have anything in it.
+
+    Counting publishable schools instead let /schools/kent/primary publish
+    with none of its five rows carrying a result, and left 44 phase pages
+    majority-blank. It is the same rule as "no page without a local average",
+    which was never extended per phase.
+
+    All-through schools count toward both phases, matching the PHASE_GROUPS
+    mapping the search filters already use.
+    """
     from backend.app import PHASE_GROUPS
 
     if "phase" not in group.columns:
         return {}
     lowered = group["phase"].fillna("").str.lower()
+
     out: dict[str, tuple[int, ...]] = {}
     for phase in ("primary", "secondary"):
         wanted = PHASE_GROUPS.get(phase, set())
         subset = group[lowered.isin(wanted)]
+
+        # The page lists every school of the phase; the threshold counts only
+        # those carrying a result, so a mostly-empty table never publishes.
+        metric = _PHASE_METRIC[phase]
+        with_result = (
+            {int(u) for u in subset.loc[subset[metric].notna(), "urn"]}
+            if metric in subset.columns else set()
+        )
+        if len(with_result & publishable) < MIN_SCHOOLS:
+            continue
+
         urns = tuple(sorted({int(u) for u in subset["urn"]} & publishable))
         if urns:
             out[phase] = urns
@@ -191,7 +224,14 @@ def _outcode(postcode) -> str | None:
 def _outcode_places(df, publishable: set[int]) -> dict[str, Place]:
     """One Place per postcode district clearing the threshold.
 
-    These carry no phase variants: nobody searches "primary schools in SW11".
+    These carry no phase variants: nobody searches "primary schools in SW11",
+    so the spec gives them no /primary or /secondary route. `phase_urns` is
+    left empty rather than computed and then filtered downstream — the
+    registry is the one place that decides which phases a place publishes,
+    and the page links whatever it reports.
+
+    Computing them here put a link to a route that does not exist on every one
+    of the 1,720 outcode pages.
     """
     if "postcode" not in df.columns:
         return {}
@@ -206,8 +246,7 @@ def _outcode_places(df, publishable: set[int]) -> dict[str, Place]:
         authorities = _authorities(group)
         place = Place(kind="outcode", slug=str(oc).lower(), name=str(oc),
                       urns=urns, parent_authority=_parent_authority(authorities),
-                      authorities=authorities,
-                      phase_urns=_phase_urns(group, publishable))
+                      authorities=authorities)
         out[place.key] = place
     return out
 

backend/tests/test_destinations_api.py

@@ -0,0 +1,269 @@
+"""The destinations serialiser's contract.
+
+Not rendering a figure is not the same as not publishing it. This endpoint is
+public and unauthenticated, so whatever the payload carries is published,
+whatever the UI draws. The categories sum to the cohort and the pupil groups
+sum to each other, so a lone suppressed cell is solvable by subtraction — the
+serialiser adds secondary suppression to prevent it.
+
+See docs/superpowers/specs/2026-08-28-destination-measures-design.md.
+"""
+
+from backend.data_loader import (
+    _destinations_block, _format_cohort_year, disclosure_invariant_holds,
+)
+
+
+def _row(group, measure, pupils, status, cohort=180, percentage=None, year=202223):
+    return {
+        "pupil_group": group,
+        "destination_measure": measure,
+        "pupils": pupils,
+        "percentage": percentage,
+        "status": status,
+        "cohort_pupils": cohort,
+        "year": year,
+    }
+
+
+def test_suppressed_category_serialises_as_suppressed_with_null_pupils():
+    rows = [
+        _row("all", "school_sixth_form", 75, "published", percentage=41.7),
+        _row("all", "sixth_form_college", None, "suppressed"),
+    ]
+    block = _destinations_block(rows)
+    cats = {c["category"]: c for c in block["groups"]["all"]["categories"]}
+    assert cats["sixth_form_college"]["status"] == "suppressed"
+    assert cats["sixth_form_college"]["pupils"] is None
+    assert cats["sixth_form_college"]["percentage"] is None
+
+
+def test_published_category_keeps_its_figures():
+    block = _destinations_block([
+        _row("all", "school_sixth_form", 75, "published", percentage=41.7),
+    ])
+    cat = block["groups"]["all"]["categories"][0]
+    assert cat["pupils"] == 75
+    assert cat["percentage"] == 41.7
+    assert cat["status"] == "published"
+
+
+def test_only_the_latest_year_is_served():
+    rows = [
+        _row("all", "school_sixth_form", 60, "published", year=202122),
+        _row("all", "school_sixth_form", 75, "published", year=202223),
+    ]
+    block = _destinations_block(rows)
+    assert block["cohort_year"] == "2022/23"
+    assert len(block["groups"]["all"]["categories"]) == 1
+    assert block["groups"]["all"]["categories"][0]["pupils"] == 75
+
+
+def test_all_three_pupil_groups_are_carried():
+    rows = [
+        _row("all", "school_sixth_form", 75, "published"),
+        _row("disadvantaged", "school_sixth_form", 17, "published", cohort=62),
+        _row("other", "school_sixth_form", 58, "published", cohort=118),
+    ]
+    block = _destinations_block(rows)
+    assert set(block["groups"]) == {"all", "disadvantaged", "other"}
+    assert block["groups"]["disadvantaged"]["cohort"] == 62
+
+
+def test_cohort_year_is_reported_so_the_page_can_date_itself():
+    block = _destinations_block([_row("all", "school_sixth_form", 75, "published")])
+    assert block["cohort_year"] == "2022/23"
+
+
+def test_format_cohort_year_handles_the_six_digit_form():
+    assert _format_cohort_year(202223) == "2022/23"
+    assert _format_cohort_year(None) is None
+
+
+def test_empty_rows_yield_none_not_an_empty_shell():
+    assert _destinations_block([]) is None
+
+
+# ── Disclosure control ──────────────────────────────────────────────────────
+#
+# The rendering guards in lib/destinations.ts stop a withheld figure being
+# DRAWN. They do nothing about it being COMPUTED: this endpoint is public and
+# unauthenticated, so whatever the payload carries is published. These tests
+# are the ones that matter.
+
+def _solve_residual(group):
+    """What any caller can work out: cohort minus everything published."""
+    published = [c["pupils"] for c in group["categories"] if c["pupils"] is not None]
+    hidden = [c for c in group["categories"] if c["status"] == "suppressed"]
+    return group["cohort"] - sum(published), len(hidden)
+
+
+def test_a_lone_suppressed_category_cannot_be_solved_for():
+    """Whitley Bay High School's real 2022/23 disadvantaged group: further
+    education withheld, everything else published, cohort 41. Before secondary
+    suppression the payload gave the answer away as 41 - 23 = 18."""
+    rows = [
+        _row("disadvantaged", "school_sixth_form", 15, "published", cohort=41),
+        _row("disadvantaged", "sixth_form_college", 0, "published", cohort=41),
+        _row("disadvantaged", "further_education", None, "suppressed", cohort=41),
+        _row("disadvantaged", "apprenticeship", 1, "published", cohort=41),
+        _row("disadvantaged", "employment", 2, "published", cohort=41),
+        _row("disadvantaged", "not_sustained", 3, "published", cohort=41),
+        _row("disadvantaged", "not_captured", 2, "published", cohort=41),
+    ]
+    group = _destinations_block(rows)["groups"]["disadvantaged"]
+    residual, hidden = _solve_residual(group)
+    assert hidden >= 2, "a lone suppressed cell must gain a companion"
+    assert residual != 18, "the withheld figure is recoverable from the payload"
+
+
+def test_every_group_hides_none_or_at_least_two_categories():
+    rows = [
+        _row("all", "school_sixth_form", 75, "published"),
+        _row("all", "sixth_form_college", None, "suppressed"),
+        _row("all", "further_education", 61, "published"),
+        _row("all", "apprenticeship", 8, "published"),
+        _row("all", "employment", 6, "published"),
+        _row("all", "not_sustained", 5, "published"),
+        _row("all", "not_captured", 4, "published"),
+    ]
+    group = _destinations_block(rows)["groups"]["all"]
+    hidden = [c for c in group["categories"] if c["status"] == "suppressed"]
+    assert len(hidden) >= 2
+
+
+def test_a_category_hidden_in_one_group_is_hidden_in_a_second():
+    """disadvantaged + other = all for every category, so a category withheld
+    in exactly one of the three is recoverable from the other two."""
+    rows = []
+    for measure, a, d, o in [
+        ("school_sixth_form", 75, None, 58),
+        ("further_education", 61, 27, 34),
+        ("apprenticeship", 8, 4, 4),
+        ("employment", 6, 1, 5),
+        ("not_sustained", 5, 3, 2),
+        ("not_captured", 4, 2, 2),
+    ]:
+        rows.append(_row("all", measure, a, "published", cohort=159))
+        rows.append(_row("disadvantaged", measure, d,
+                         "published" if d is not None else "suppressed", cohort=37))
+        rows.append(_row("other", measure, o, "published", cohort=122))
+
+    groups = _destinations_block(rows)["groups"]
+    measures = {c["category"] for g in groups.values() for c in g["categories"]}
+    assert len(measures) == 6, "the fixture's six measures must all be checked"
+
+    for measure in sorted(measures):
+        hidden = sum(
+            1 for g in groups.values() for c in g["categories"]
+            if c["category"] == measure and c["status"] == "suppressed"
+        )
+        # The invariant is "none, or at least two" — not "at least two".
+        assert hidden != 1, f"{measure} is solvable across the pupil groups"
+
+
+def test_a_suppressed_cell_never_keeps_its_percentage():
+    """percentage / pupils would hand back the cohort, and with it the residual."""
+    rows = [
+        _row("all", "school_sixth_form", 75, "published", percentage=41.7),
+        _row("all", "sixth_form_college", None, "suppressed", percentage=11.7),
+        _row("all", "further_education", 61, "published", percentage=33.9),
+    ]
+    group = _destinations_block(rows)["groups"]["all"]
+    for cell in group["categories"]:
+        if cell["status"] != "published":
+            assert cell["pupils"] is None
+            assert cell["percentage"] is None
+
+
+def test_aggregates_are_not_served():
+    """An aggregate spanning exactly one suppressed component names it, and
+    nothing renders them today."""
+    rows = [
+        _row("all", "school_sixth_form", 75, "published"),
+        _row("all", "agg_sustained_all", 171, "published"),
+    ]
+    group = _destinations_block(rows)["groups"]["all"]
+    assert [c["category"] for c in group["categories"]] == ["school_sixth_form"]
+    assert "aggregates" not in group
+
+
+def test_a_fully_published_group_is_left_alone():
+    """Secondary suppression must not cost anything where nothing is withheld —
+    this is the all-pupils view on every mainstream secondary."""
+    rows = [
+        _row("all", m, p, "published")
+        for m, p in [("school_sixth_form", 75), ("sixth_form_college", 21),
+                     ("further_education", 61), ("apprenticeship", 8),
+                     ("employment", 6), ("not_sustained", 5), ("not_captured", 4)]
+    ]
+    group = _destinations_block(rows)["groups"]["all"]
+    assert all(c["status"] == "published" for c in group["categories"])
+    assert len(group["categories"]) == 7
+
+
+def test_the_invariant_is_asserted_directly_not_re_derived():
+    """A group with one suppressed category and nothing else to withhold."""
+    rows = [
+        _row("all", "school_sixth_form", None, "suppressed", cohort=9),
+        _row("all", "sixth_form_college", None, "not_applicable", cohort=9),
+        _row("all", "further_education", None, "not_applicable", cohort=9),
+    ]
+    block = _destinations_block(rows)
+    assert block is None or disclosure_invariant_holds(block["groups"])
+
+
+def test_a_sparse_cohort_with_no_companion_drops_the_group():
+    """Special schools and AP routinely have one suppressed category and every
+    other one not applicable. There is nothing left to withhold, so the group
+    goes — an earlier version returned here with the violation intact."""
+    rows = [
+        _row("all", "school_sixth_form", None, "suppressed", cohort=9),
+        _row("all", "sixth_form_college", None, "not_applicable", cohort=9),
+        _row("all", "further_education", None, "not_applicable", cohort=9),
+        _row("all", "apprenticeship", None, "not_applicable", cohort=9),
+        _row("all", "employment", None, "not_applicable", cohort=9),
+        _row("all", "not_sustained", None, "not_applicable", cohort=9),
+        _row("all", "not_captured", None, "not_applicable", cohort=9),
+    ]
+    block = _destinations_block(rows)
+    assert block is None or "all" not in block["groups"], (
+        "a group that cannot be made safe must not be served"
+    )
+
+
+def test_zeros_are_not_treated_as_a_usable_companion():
+    """Suppressing a zero protects nothing — the residual is unchanged. With
+    only zeros available the group must be dropped, not falsely 'fixed'."""
+    rows = [
+        _row("all", "school_sixth_form", None, "suppressed", cohort=5),
+        _row("all", "sixth_form_college", 0, "published", cohort=5),
+        _row("all", "further_education", 0, "published", cohort=5),
+    ]
+    block = _destinations_block(rows)
+    if block and "all" in block["groups"]:
+        group = block["groups"]["all"]
+        published = sum(c["pupils"] for c in group["categories"]
+                        if c["pupils"] is not None)
+        hidden = [c for c in group["categories"] if c["status"] == "suppressed"]
+        assert len(hidden) != 1, "a zero companion leaves the figure solvable"
+        assert group["cohort"] - published != 5
+
+
+def test_masking_always_terminates_in_a_safe_state():
+    """Exhaustive over every suppression pattern of a four-category group."""
+    from itertools import product
+    MEASURES = ["school_sixth_form", "sixth_form_college",
+                "further_education", "apprenticeship"]
+    for statuses in product(["published", "suppressed", "not_applicable"],
+                            repeat=len(MEASURES)):
+        rows = [
+            _row("all", m, 3 if st == "published" else None, st, cohort=12)
+            for m, st in zip(MEASURES, statuses)
+        ]
+        block = _destinations_block(rows)
+        if block is None:
+            continue
+        assert disclosure_invariant_holds(block["groups"]), (
+            f"invariant broken for {statuses}"
+        )

backend/tests/test_flags.py

@@ -0,0 +1,163 @@
+"""Tests for the feature flag layer (spec 2026-08-23).
+
+None of these need a running Unleash. That is the point: an unset UNLEASH_URL
+means every flag is False, which is what local development and CI get.
+"""
+
+from datetime import date, timedelta
+
+from backend import flags
+
+
+def test_every_declared_flag_is_keyed_by_its_own_name():
+    # One string is the registry key, the Unleash flag name and the JSON key.
+    # A mismatch here would mean the UI toggles a flag the code never reads.
+    for key, flag in flags.REGISTRY.items():
+        assert key == flag.name
+
+
+def test_flag_names_are_snake_case():
+    # Matches the API's existing convention (admission_distance,
+    # rwm_expected_pct) so no case transformation exists to get wrong.
+    for name in flags.REGISTRY:
+        assert name == name.lower()
+        assert "-" not in name and " " not in name
+
+
+def test_an_unconfigured_client_evaluates_every_flag_false(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    for name in flags.REGISTRY:
+        assert flags.is_enabled(name) is False
+
+
+def test_an_undeclared_flag_is_false_rather_than_an_error(monkeypatch):
+    # A typo'd flag name must not raise in a request path. It is logged as an
+    # error, because an undeclared flag is always a bug.
+    monkeypatch.setattr(flags, "_client", None)
+    assert flags.is_enabled("no_such_flag") is False
+
+
+def test_an_exploding_client_is_false_rather_than_a_500(monkeypatch):
+    class Boom:
+        def is_enabled(self, *a, **kw):
+            raise RuntimeError("unleash is on fire")
+
+    monkeypatch.setattr(flags, "_client", Boom())
+    name = next(iter(flags.REGISTRY))
+    assert flags.is_enabled(name) is False
+
+
+def test_all_flags_reports_every_declared_flag(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    assert set(flags.all_flags()) == set(flags.REGISTRY)
+    assert all(v is False for v in flags.all_flags().values())
+
+
+def test_a_flag_older_than_the_limit_fails_this_test():
+    """A tripwire, not an assertion about correctness.
+
+    Flags are temporary scaffolding and the failure mode of every flag system
+    is accumulation. This fails on the day a flag turns 90, on whatever PR
+    happens to be open — which is the point: someone has to decide.
+
+    To fix: delete the flag and the branches that read it, or, if it genuinely
+    still needs to exist, move its `added` date and say why in the commit.
+    """
+    stale = [
+        f.name for f in flags.REGISTRY.values()
+        if date.today() - f.added > timedelta(days=flags.MAX_FLAG_AGE_DAYS)
+    ]
+    assert not stale, (
+        f"Flags older than {flags.MAX_FLAG_AGE_DAYS} days: {stale}. "
+        "Remove the flag and the code branches it guards, or move its `added` "
+        "date deliberately."
+    )
+
+
+def _client():
+    from fastapi.testclient import TestClient
+    from backend import app as app_module
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_the_flags_endpoint_lists_every_declared_flag(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    body = _client().get("/api/flags").json()
+    assert set(body) == set(flags.REGISTRY)
+
+
+def test_the_flags_endpoint_answers_false_when_unleash_is_unreachable(monkeypatch):
+    # The endpoint must still answer. A frontend that cannot read flags renders
+    # everything dark, which is right; one that gets a 500 renders nothing.
+    monkeypatch.setattr(flags, "_client", None)
+    res = _client().get("/api/flags")
+    assert res.status_code == 200
+    assert all(v is False for v in res.json().values())
+
+
+def _school_payload(monkeypatch, *, flag_on: bool):
+    """Fetch one school's payload with the distance flag forced on or off.
+
+    The DataFrame shape is copied from test_school_details.py rather than
+    minimised: the endpoint reads a wide set of GIAS columns, and a trimmed
+    frame fails for reasons that have nothing to do with flags.
+    """
+    import numpy as np
+    import pandas as pd
+    from fastapi.testclient import TestClient
+    from backend import app as app_module
+
+    df = pd.DataFrame([{
+        "urn": 150275,
+        "school_name": "West London Performing Arts Academy",
+        "phase": "Secondary",
+        "school_type": "Special post 16 institution",
+        "trust_name": None,
+        "religious_denomination": "Does not apply",
+        "gender": None,
+        "age_range": "16-25",
+        "admissions_policy": None,
+        "capacity": np.nan,
+        "gias_total_pupils": np.nan,
+        "headteacher_name": None,
+        "website": None,
+        "ofsted_grade": np.nan,
+        "local_authority": "Ealing",
+        "address": "268 Northfield Avenue, London, W5 4UB",
+        "postcode": "W5 4UB",
+        "latitude": 51.4986,
+        "longitude": -0.3148,
+        "year": np.nan,
+        "total_pupils": np.nan,
+        "eligible_pupils": np.nan,
+        "rwm_expected_pct": np.nan,
+    }])
+
+    monkeypatch.setattr(app_module, "load_school_data", lambda: df)
+    # Two arguments: get_supplementary_data(db, urn). See backend/app.py.
+    monkeypatch.setattr(
+        app_module, "get_supplementary_data",
+        lambda db, urn: {"admission_distance": {"distance_m": 772.49,
+                                                "year": 2024}})
+    monkeypatch.setattr(flags, "is_enabled", lambda name: flag_on)
+
+    client = TestClient(app_module.app, raise_server_exceptions=False)
+    res = client.get("/api/schools/150275")
+    assert res.status_code == 200, res.text
+    return res.json()
+
+
+def test_the_distance_field_is_absent_when_the_flag_is_off(monkeypatch):
+    """Absent, not null, and withheld at the source.
+
+    /api/schools/ is public and unauthenticated. Leaving a withheld field in
+    the payload while declining to render it hands the record to anyone who
+    opens the network tab — the reasoning already recorded in c9a1892.
+    """
+    body = _school_payload(monkeypatch, flag_on=False)
+    assert "admission_distance" not in body
+
+
+def test_the_distance_field_is_present_when_the_flag_is_on(monkeypatch):
+    body = _school_payload(monkeypatch, flag_on=True)
+    assert body["admission_distance"]["distance_m"] == 772.49

backend/tests/test_places.py

@@ -343,3 +343,78 @@ def test_the_merged_place_takes_its_most_common_spelling():
             + _town(5, "NEWCASTLE-UNDER-LYME", "Staffordshire", start=400000))
     reg = build_place_registry(_df(rows))
     assert reg["town:newcastle-under-lyme"].name == "Newcastle-under-Lyme"
+
+
+def test_a_phase_page_needs_results_not_merely_publishable_schools():
+    """/schools/kent/primary published with none of its five rows scored.
+
+    The threshold counted schools that were publishable — a result OR an
+    Ofsted grade — while the page exists for its results column. Forty-four
+    phase pages were majority-blank; one had no results at all.
+    """
+    rows = _town(MIN_SCHOOLS, "Kent", "Kent")
+    for r in rows:
+        r["rwm_expected_pct"] = np.nan     # Ofsted only, no results
+    reg = build_place_registry(_df(rows))
+
+    assert "town:kent" in reg               # the place still publishes
+    assert not reg["town:kent"].publishes_phase("primary")
+
+
+def test_a_phase_page_publishes_once_enough_schools_carry_a_result():
+    rows = _town(MIN_SCHOOLS, "Beccles", "Suffolk")
+    reg = build_place_registry(_df(rows))
+    assert reg["town:beccles"].publishes_phase("primary")
+
+
+def test_a_publishing_phase_page_still_lists_its_unscored_schools():
+    """The threshold gates whether the page exists; it does not filter rows.
+
+    A parent looking up a school by name has to find it whether or not it
+    published results.
+    """
+    scored = _town(MIN_SCHOOLS, "Beccles", "Suffolk", start=300000)
+    unscored = _town(2, "Beccles", "Suffolk", start=400000)
+    for r in unscored:
+        r["rwm_expected_pct"] = np.nan
+    reg = build_place_registry(_df(scored + unscored))
+
+    place = reg["town:beccles"]
+    assert place.publishes_phase("primary")
+    assert len(place.phase_urns["primary"]) == MIN_SCHOOLS + 2
+
+
+def test_the_secondary_threshold_counts_its_own_metric():
+    # A town full of scored primaries must not thereby publish a secondary page.
+    rows = _town(MIN_SCHOOLS, "Brentwood", "Essex")
+    reg = build_place_registry(_df(rows))
+    assert not reg["town:brentwood"].publishes_phase("secondary")
+
+
+def test_an_outcode_publishes_no_phase_variants():
+    """There is no /schools/near/[outcode]/[phase] route, by design.
+
+    Nobody searches "primary schools in SW11", so the spec gives outcodes no
+    phase variants. The registry computed them anyway, and the place page —
+    which links whatever phases the registry reports — put two 404s on every
+    outcode page in the site.
+
+    This is the single rule now: a kind with no phase route reports no phases,
+    so neither the page nor the sitemap can offer one.
+    """
+    rows = [{"urn": 500000 + i, "school_name": f"SW11 School {i}",
+             "town": "London", "local_authority": "Wandsworth",
+             "postcode": "SW11 1AA"} for i in range(MIN_SCHOOLS + 3)]
+    reg = build_place_registry(_df(rows))
+
+    place = reg["outcode:sw11"]
+    assert place.phase_urns == {}
+    assert not place.publishes_phase("primary")
+    assert not place.publishes_phase("secondary")
+
+
+def test_an_authority_still_publishes_phase_variants():
+    """Authorities keep theirs — "primary schools in Kent" is a real query,
+    and /schools/authority/[la]/[phase] is the route that serves it."""
+    reg = build_place_registry(_df(_town(MIN_SCHOOLS, "Maidstone", "Kent")))
+    assert reg["authority:kent"].publishes_phase("primary")

backend/tests/test_places_api.py

@@ -45,10 +45,30 @@ def test_registry_carries_a_count_per_place(client):
     assert town["count"] == 6
 
 
-def test_place_detail_returns_its_schools_ranked(client):
+def test_place_detail_returns_its_schools_alphabetically(client):
+    """A place page is read by someone looking for a school they can name.
+
+    Scanning for it is what the order should serve, so the list is A-Z.
+    /api/rankings is where the league-table ordering lives.
+    """
     body = client.get("/api/places/town/brentwood").json()
     assert body["place"]["name"] == "Brentwood"
-    scores = [s["rwm_expected_pct"] for s in body["schools"]]
+    names = [s["school_name"] for s in body["schools"]]
+    assert names == sorted(names, key=str.lower)
+
+
+def test_place_ordering_ignores_case(client):
+    body = client.get("/api/places/town/brentwood").json()
+    names = [s["school_name"] for s in body["schools"]]
+    # A capitalised name must not sort ahead of every lowercase one.
+    assert names == sorted(names, key=str.lower)
+
+
+def test_the_rankings_endpoint_still_ranks_by_metric(client):
+    # Alphabetical is a place-page decision, not a site-wide one.
+    body = client.get("/api/rankings?metric=rwm_expected_pct&phase=primary").json()
+    scores = [r["rwm_expected_pct"] for r in body.get("rankings", [])
+              if r.get("rwm_expected_pct") is not None]
     assert scores == sorted(scores, reverse=True)
 
 
@@ -69,3 +89,97 @@ def test_unknown_place_404s(client):
 
 def test_unknown_kind_404s(client):
     assert client.get("/api/places/planet/mars").status_code == 404
+
+
+def _straddling_df() -> pd.DataFrame:
+    """Eight schools in CM13: six in Essex, which has a page, and two in an
+    authority too small to have one.
+
+    Two, not one: the registry ignores an authority holding a single school in
+    a place, because GIAS carries occasional postcode errors."""
+    df = _schools_df()
+    extra = df.iloc[:2].copy()
+    extra["urn"] = [200000, 200001]
+    extra["school_name"] = ["Scilly School 0", "Scilly School 1"]
+    extra["local_authority"] = "Isles Of Scilly"
+    return pd.concat([df, extra], ignore_index=True)
+
+
+@pytest.fixture()
+def straddling_client(monkeypatch):
+    from backend import app as app_module
+
+    monkeypatch.setattr(app_module, "load_school_data", _straddling_df)
+    monkeypatch.setattr(app_module, "load_latest_school_data", _straddling_df)
+    monkeypatch.setattr(app_module, "_place_registry", None)
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_an_outcode_reports_no_phases_because_it_has_no_phase_route(client):
+    body = client.get("/api/places/outcode/cm13").json()
+    assert body["place"]["phases"] == []
+
+
+def test_an_authority_reports_the_phases_it_publishes(client):
+    body = client.get("/api/places/authority/essex").json()
+    assert body["place"]["phases"] == ["primary"]
+
+
+def test_an_authority_without_a_page_is_named_but_carries_no_slug(straddling_client):
+    """Two English authorities — City of London and the Isles of Scilly — hold
+    fewer than the five schools a page needs, so they have no page.
+
+    Naming them is still right: the page says where the place is. Linking them
+    would not be. A null slug is what tells the page to print the name plainly
+    rather than invent a URL that 404s.
+    """
+    body = straddling_client.get("/api/places/outcode/cm13").json()
+    by_name = {a["name"]: a for a in body["place"]["authorities"]}
+    assert by_name["Essex"]["slug"] == "essex"
+    assert by_name["Isles Of Scilly"]["slug"] is None
+
+
+def _attributed_df() -> pd.DataFrame:
+    """The same town, with the four attributes the place table now shows."""
+    df = _schools_df()
+    df["age_range"] = "4-11"
+    df["religious_denomination"] = "Church of England"
+    df["nursery_provision"] = True
+    df["parliamentary_constituency"] = "Brentwood and Ongar"
+    return df
+
+
+@pytest.fixture()
+def attributed_client(monkeypatch):
+    from backend import app as app_module
+
+    monkeypatch.setattr(app_module, "load_school_data", _attributed_df)
+    monkeypatch.setattr(app_module, "load_latest_school_data", _attributed_df)
+    monkeypatch.setattr(app_module, "_place_registry", None)
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_place_detail_carries_the_attributes_the_table_shows(attributed_client):
+    """age_range and religious_denomination ride in on SCHOOL_COLUMNS.
+
+    nursery_provision and parliamentary_constituency do not, and the place
+    table needs all four — a column the response cannot fill is a column of
+    dashes on ~3,900 pages.
+    """
+    body = attributed_client.get("/api/places/town/brentwood").json()
+    school = body["schools"][0]
+    assert school["age_range"] == "4-11"
+    assert school["religious_denomination"] == "Church of England"
+    assert school["nursery_provision"] is True
+    assert school["parliamentary_constituency"] == "Brentwood and Ongar"
+
+
+def test_place_detail_survives_a_mart_without_the_optional_columns(client):
+    """The base fixture has neither column, as an unrebuilt mart does not.
+
+    data_loader degrades those to NULL rather than failing the load, so the
+    endpoint must not assume they are present.
+    """
+    res = client.get("/api/places/town/brentwood")
+    assert res.status_code == 200
+    assert "nursery_provision" not in res.json()["schools"][0]

backend/tests/test_rate_limit_key.py

@@ -0,0 +1,147 @@
+"""The rate-limit bucket must be the caller, not the proxy in front of them.
+
+`get_remote_address` reads request.client.host. In staging and production the
+backend has no published ports and its only caller is the Next proxy, so that
+host is the Next container — one bucket for every browser user on the site.
+Measured before this fix: 70 concurrent requests, 60 served and 10 refused.
+"""
+
+from starlette.datastructures import Headers
+
+from backend.app import client_key
+
+
+class _Req:
+    """Enough of a Request for the key function: headers and a client host."""
+
+    def __init__(self, headers: dict, host: str = "10.0.0.9"):
+        self.headers = Headers(headers)
+        self.client = type("C", (), {"host": host})()
+        self.scope = {"type": "http", "client": (host, 0),
+                      "headers": [(k.lower().encode(), v.encode())
+                                  for k, v in headers.items()]}
+
+
+def test_cloudflare_header_wins():
+    # Cloudflare sets CF-Connecting-IP and overwrites any client-supplied
+    # value, so it is trustworthy in a way a parsed XFF chain is not.
+    assert client_key(_Req({"cf-connecting-ip": "203.0.113.7"})) == "203.0.113.7"
+
+
+def test_forwarded_for_is_the_fallback_and_takes_the_first_entry():
+    # Left-most is the original client; everything after it is proxies.
+    assert client_key(
+        _Req({"x-forwarded-for": "203.0.113.7, 10.0.0.2"})) == "203.0.113.7"
+
+
+def test_remote_address_is_the_last_resort():
+    assert client_key(_Req({}, host="10.0.0.9")) == "10.0.0.9"
+
+
+def test_cloudflare_header_beats_forwarded_for():
+    key = client_key(_Req({"cf-connecting-ip": "203.0.113.7",
+                           "x-forwarded-for": "198.51.100.1"}))
+    assert key == "203.0.113.7"
+
+
+def test_two_callers_get_two_buckets():
+    # The whole point: one user exhausting their limit must not refuse another.
+    a = client_key(_Req({"cf-connecting-ip": "203.0.113.7"}))
+    b = client_key(_Req({"cf-connecting-ip": "203.0.113.8"}))
+    assert a != b
+
+
+def test_whitespace_is_stripped():
+    # "a, b" split on comma leaves a leading space on every entry but the
+    # first; an unstripped key silently creates a second bucket per client.
+    assert client_key(_Req({"x-forwarded-for": " 203.0.113.7 ,10.0.0.2"})) \
+        == "203.0.113.7"
+
+
+# ---------------------------------------------------------------------------
+# The ceiling that header rotation cannot raise.
+# ---------------------------------------------------------------------------
+
+import pytest
+from fastapi.testclient import TestClient
+
+
+@pytest.fixture()
+def api(monkeypatch):
+    from backend import app as app_module
+    from backend.config import settings
+
+    monkeypatch.setattr(settings, "global_rate_limit_per_minute", 5)
+    monkeypatch.setattr(app_module, "_global_window", None)
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+
+def _ceiling_req(path: str, host: str):
+    """Enough of a Request for exempt_from_ceiling: a path and a peer host."""
+    return type("R", (), {
+        "url": type("U", (), {"path": path})(),
+        "client": type("C", (), {"host": host})(),
+    })()
+
+
+def _get(client, path="/api/flags", cf=None):
+    headers = {"cf-connecting-ip": cf} if cf else {}
+    return client.get(path, headers=headers)
+
+
+def test_rotating_the_cloudflare_header_cannot_buy_unlimited_requests(api):
+    """The attack the per-client keying opened up.
+
+    client_key trusts CF-Connecting-IP, and nothing in this process can tell an
+    edge-set header from an attacker-set one — that distinction can only be
+    made at Cloudflare, with Authenticated Origin Pulls or an origin firewall.
+    A caller reaching the origin directly can therefore mint a fresh
+    rate-limit bucket per request and evade per-client limits entirely.
+
+    Per-client fairness is still the right default; this is the backstop that
+    bounds what evading it can achieve. Without it, correct keying would be a
+    net regression against abuse compared with the shared bucket it replaced.
+    """
+    codes = [_get(api, cf=f"203.0.113.{i}").status_code for i in range(8)]
+    assert codes.count(200) == 5
+    assert codes.count(429) == 3
+
+
+def test_the_ceiling_says_which_limit_was_hit(api):
+    # Distinguishable from slowapi's per-client 429, or an operator reading
+    # logs cannot tell "one noisy client" from "the origin is saturated".
+    for i in range(5):
+        _get(api, cf=f"203.0.113.{i}")
+    refused = _get(api, cf="203.0.113.99")
+    assert refused.status_code == 429
+    assert "capacity" in refused.json()["detail"].lower()
+    assert refused.headers.get("retry-after")
+
+
+def test_traffic_below_the_ceiling_is_untouched(api):
+    codes = [_get(api, cf=f"203.0.113.{i}").status_code for i in range(5)]
+    assert codes == [200] * 5
+
+
+def test_the_container_healthcheck_is_exempt(api):
+    """The healthcheck runs `curl http://localhost:80/api/data-info` inside the
+    container. If the ceiling could starve it, saturation would fail the
+    healthcheck, restart the container, and turn a load spike into an outage
+    loop — the ceiling has to protect the origin, not kill it.
+    """
+    from backend.app import exempt_from_ceiling
+
+    assert exempt_from_ceiling(_ceiling_req("/api/data-info", "127.0.0.1"))
+    assert exempt_from_ceiling(_ceiling_req("/api/data-info", "::1"))
+    # Everyone else is counted.
+    assert not exempt_from_ceiling(_ceiling_req("/api/data-info", "10.0.0.9"))
+
+
+def test_the_ceiling_ignores_non_api_paths():
+    # Sitemaps and robots.txt are served by this app too, and a crawler
+    # fetching them must not be refused because the API is busy.
+    from backend.app import exempt_from_ceiling
+
+    assert exempt_from_ceiling(_ceiling_req("/sitemap.xml", "10.0.0.9"))
+    assert exempt_from_ceiling(_ceiling_req("/robots.txt", "10.0.0.9"))

backend/tests/test_sitemap.py

@@ -288,3 +288,19 @@ def test_outcodes_get_no_phase_variants(place_sitemaps):
     # Nobody searches "primary schools in CM13"; the routes do not exist.
     xml = place_sitemaps["outcodes-1.xml"]
     assert "/primary" not in xml and "/secondary" not in xml
+
+
+def test_authority_phase_variants_are_submitted_in_their_own_namespace(place_sitemaps):
+    """302 of these were already in the sitemap, and every one 404'd.
+
+    The spec gives authorities a phase route; the plan built the bare
+    authority route and dropped it. Nothing noticed because the sitemap was
+    written from the registry, which was right, while the routes were written
+    by hand. This test fails if the URL ever leaves the sitemap; the e2e
+    journey fails if the route ever leaves the app.
+    """
+    xml = place_sitemaps["places-1.xml"]
+    assert ("<loc>https://www.schoolcompare.co.uk"
+            "/schools/authority/essex/primary</loc>") in xml
+    # And never in the town namespace, which is a different set of schools.
+    assert "/schools/essex/primary" not in xml

backend/tests/test_suggest.py

@@ -0,0 +1,157 @@
+"""Tests for school autosuggest (spec 2026-08-26)."""
+
+from backend import data_loader
+
+
+class _FakeDocs:
+    def __init__(self, hits, explode=False):
+        self._hits = hits
+        self._explode = explode
+        self.last_params = None
+
+    def search(self, params):
+        self.last_params = params
+        if self._explode:
+            raise RuntimeError("typesense is down")
+        return {"hits": [{"document": d} for d in self._hits]}
+
+
+class _FakeClient:
+    def __init__(self, hits, explode=False):
+        self.docs = _FakeDocs(hits, explode)
+        self.collections = {"schools": type("C", (), {"documents": self.docs})()}
+
+
+_HIT = {
+    "urn": 100010, "school_name": "Brecknock Primary School",
+    "local_authority": "Camden", "postcode": "NW1 1AA",
+    "phase": "Primary", "school_type": "Community school",
+}
+
+
+def _use(monkeypatch, client):
+    monkeypatch.setattr(data_loader, "_get_typesense_client", lambda: client)
+
+
+def test_returns_the_fields_a_suggestion_needs(monkeypatch):
+    # Local authority is not decoration: there are many schools called
+    # "St Mary's", and a list without it cannot be chosen between.
+    _use(monkeypatch, _FakeClient([_HIT]))
+    out = data_loader.suggest_schools_typesense("breck")
+    assert out == [{
+        "urn": 100010, "school_name": "Brecknock Primary School",
+        "local_authority": "Camden", "postcode": "NW1 1AA",
+        "phase": "Primary", "school_type": "Community school",
+    }]
+
+
+def test_a_missing_optional_field_becomes_an_empty_string(monkeypatch):
+    # phase and school_type are optional in the Typesense schema. A missing
+    # key must not KeyError in the keystroke path.
+    _use(monkeypatch, _FakeClient([{"urn": 1, "school_name": "X",
+                                    "local_authority": "Y", "postcode": "Z"}]))
+    out = data_loader.suggest_schools_typesense("x")
+    assert out[0]["phase"] == "" and out[0]["school_type"] == ""
+
+
+def test_typesense_unavailable_gives_no_suggestions_rather_than_raising(monkeypatch):
+    _use(monkeypatch, None)
+    assert data_loader.suggest_schools_typesense("anything") == []
+
+
+def test_a_typesense_error_gives_no_suggestions_rather_than_raising(monkeypatch):
+    _use(monkeypatch, _FakeClient([], explode=True))
+    assert data_loader.suggest_schools_typesense("anything") == []
+
+
+def test_the_limit_is_passed_through_and_clamped(monkeypatch):
+    client = _FakeClient([])
+    _use(monkeypatch, client)
+    data_loader.suggest_schools_typesense("x", limit=500)
+    assert client.docs.last_params["per_page"] == 20
+
+
+def _client(monkeypatch, rows, *, blow_up_dataframe=False):
+    from fastapi.testclient import TestClient
+    from backend import app as app_module
+
+    monkeypatch.setattr(app_module, "suggest_schools_typesense",
+                        lambda q, limit=8: rows)
+    if blow_up_dataframe:
+        def _boom():
+            raise AssertionError("the suggest path must not load the DataFrame")
+        monkeypatch.setattr(app_module, "load_school_data", _boom)
+        monkeypatch.setattr(app_module, "load_latest_school_data", _boom)
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_the_endpoint_returns_suggestions(monkeypatch):
+    body = _client(monkeypatch, [_HIT]).get("/api/suggest?q=breck").json()
+    assert body["suggestions"][0]["school_name"] == "Brecknock Primary School"
+
+
+def test_the_endpoint_never_touches_the_dataframe(monkeypatch):
+    """The whole reason this is not a mode of /api/schools.
+
+    That endpoint filters and sorts 25,000 rows of pandas per query, holding
+    the GIL. Per keystroke, that is the cost this endpoint exists to avoid.
+    """
+    res = _client(monkeypatch, [_HIT], blow_up_dataframe=True).get("/api/suggest?q=breck")
+    assert res.status_code == 200
+    assert res.json()["suggestions"]
+
+
+def test_a_one_character_query_returns_nothing_and_does_not_error(monkeypatch):
+    # The keystroke path never errors on ordinary input.
+    res = _client(monkeypatch, [_HIT]).get("/api/suggest?q=b")
+    assert res.status_code == 200
+    assert res.json() == {"suggestions": []}
+
+
+def test_a_blank_query_returns_nothing_and_does_not_error(monkeypatch):
+    res = _client(monkeypatch, [_HIT]).get("/api/suggest?q=")
+    assert res.status_code == 200
+    assert res.json() == {"suggestions": []}
+
+
+def test_typesense_down_is_an_empty_list_not_a_500(monkeypatch):
+    res = _client(monkeypatch, []).get("/api/suggest?q=breck")
+    assert res.status_code == 200
+    assert res.json() == {"suggestions": []}
+
+
+def test_the_response_is_cacheable(monkeypatch):
+    # Prefix queries repeat enormously across users, and school names change
+    # once a year. Without this the endpoint pays full price every keystroke.
+    res = _client(monkeypatch, [_HIT]).get("/api/suggest?q=breck")
+    assert "s-maxage" in res.headers.get("cache-control", "")
+    assert res.headers.get("etag")
+
+
+def test_a_malformed_urn_does_not_raise(monkeypatch):
+    """The docstring promises "never raises"; the parsing loop sat outside the
+    try, so int(None) or int("abc") would have turned a keystroke into a 500.
+
+    Typesense declares urn as int32, so this should be unreachable — but the
+    contract is what the caller relies on, and a search index is a separate
+    system that can be reindexed by something other than this code.
+    """
+    _use(monkeypatch, _FakeClient([{"urn": None, "school_name": "X",
+                                    "local_authority": "Y", "postcode": "Z"}]))
+    assert data_loader.suggest_schools_typesense("x") == []
+
+
+def test_a_malformed_row_does_not_discard_the_good_ones(monkeypatch):
+    # One bad document must not blank the whole dropdown.
+    _use(monkeypatch, _FakeClient([
+        {"urn": "not-a-number", "school_name": "Bad", "local_authority": "Y",
+         "postcode": "Z"},
+        _HIT,
+    ]))
+    out = data_loader.suggest_schools_typesense("x")
+    assert [r["urn"] for r in out] == [100010]
+
+
+def test_a_hit_with_no_document_does_not_raise(monkeypatch):
+    _use(monkeypatch, _FakeClient([{}]))
+    assert data_loader.suggest_schools_typesense("x") == []

backend/tests/test_supplementary_batch.py

@@ -132,16 +132,23 @@ def test_one_query_per_table_and_latest_row_per_urn():
         "FactPupilCharacteristics": [],
         "FactDeprivation": [],
         "FactFinance": [],
+        "FactKs4Destinations": [],
+        "FactKs5Destinations": [],
     }
     session = _FakeSession(rows)
     out = get_supplementary_data_batch(session, [1, 2])
 
-    # Exactly one query per table — six total, regardless of two URNs.
+    # Exactly one query per table — eight total, regardless of two URNs.
     assert sorted(session.queries) == [
         "FactAdmissionDistance", "FactAdmissions", "FactDeprivation",
-        "FactFinance", "FactOfstedInspection", "FactPupilCharacteristics",
+        "FactFinance", "FactKs4Destinations", "FactKs5Destinations",
+        "FactOfstedInspection", "FactPupilCharacteristics",
     ]
 
+    # A school with no destination rows gets null, not an empty shell — the
+    # frontend renders the section from the block's presence.
+    assert out[1]["destinations"] is None
+
     # Latest Ofsted kept per URN
     assert out[1]["ofsted"]["overall_effectiveness"] == 2
     assert out[2]["ofsted"]["overall_effectiveness"] == 1

claude.md

@@ -23,6 +23,52 @@ Key files:
 - `backend/data_loader.py` - Data queries, geocoding, legacy DataFrame compatibility
 - `backend/schemas.py` - Column mappings, metric definitions, LA code mappings
 
+### Content / CMS (Payload)
+
+Payload CMS runs **inside** the Next.js app — one image, one container, no
+separate service. It powers `/blog`; `/about` is a plain coded page.
+
+- **Admin panel:** `/admin`. The only authenticated surface on the site.
+  `noindex` via both `robots.txt` and `X-Robots-Tag`.
+- **CMS API:** `/cms-api`, **not** `/api`. `/api/*` is a catch-all proxy to
+  FastAPI (`app/(frontend)/api/[...path]`) which would silently swallow every
+  admin call and forward it to the backend. Mount points are defined once in
+  `lib/payloadRoutes.ts`.
+- **Database:** the existing Postgres, in its own `payload` schema, so no
+  pipeline operation on `public` — including
+  `scripts/migrate_csv_to_db.py --drop` — can reach blog content.
+- **Uploads:** the `payload_media` Docker volume at `/app/media`. Not
+  reproducible from the pipeline; must be backed up.
+- **New env vars:** `DATABASE_URL` and `PAYLOAD_SECRET` on the frontend service.
+  Staging must use a different `PAYLOAD_SECRET` from production.
+- Publishing workflow and house style: `nextjs-app/docs/PUBLISHING.md`.
+- **Admin field components resolve through a generated import map**
+  (`app/(payload)/admin/importMap.js`). Payload hands the client a *path* per
+  field and looks it up there; a missing entry renders no field and reports no
+  error, while `required` still blocks the save. After adding or changing any
+  field, editor or lexical feature, run `npm run generate:importmap` in
+  `nextjs-app/` and commit the result.
+
+### Two route groups
+
+`nextjs-app/app/` has no root `layout.tsx`. It cannot: Payload's admin panel
+ships its own root layout rendering `<html>`/`<body>`, and Next permits
+multiple root layouts only when no `app/layout.tsx` exists.
+
+- `app/(frontend)/` — the site. Its `layout.tsx` is the site's root layout.
+- `app/(payload)/` — the admin panel and `/cms-api`.
+
+Route groups are invisible to routing, so every public URL is unchanged.
+
+**The metadata file conventions stay at the `app/` root** — `robots.ts`,
+`opengraph-image.tsx`, `icon.png`, `apple-icon.png`. Inside a route group Next
+treats them as segment-scoped: it renames `/icon.png` to `/icon-<hash>.png` and
+drops `/robots.txt` entirely. Route handlers are unaffected.
+
+The build must succeed with `DATABASE_URL` unset, because CI builds it that
+way. Never call `getCachedPayload()` at module scope, and never add
+`generateStaticParams` to a DB-backed route.
+
 ### Frontend (Vanilla JS)
 - Single-page application with hash-based routing
 - Chart.js for data visualization

docker-compose.portainer.staging.yml

@@ -16,7 +16,16 @@
 #   ADMIN_API_KEY          — Backend admin API key
 #   TYPESENSE_API_KEY      — Typesense admin API key
 #   TYPESENSE_SEARCH_KEY   — Typesense search-only key (exposed to frontend)
-#   AIRFLOW_ADMIN_USER     — Airflow admin username (password auto-generated, see api-server logs)
+#   UNLEASH_URL            — http://<unleash-ip>:4242/api (empty = all flags off)
+#   UNLEASH_API_TOKEN      — Unleash *client* token, environment: development
+#   PAYLOAD_SECRET         — Payload CMS encryption secret. REQUIRED: long and
+#                            random, and DIFFERENT from production's. Sharing
+#                            it would let a staging session authenticate
+#                            against production.
+#   AIRFLOW_ADMIN_USER     — Airflow admin username (default: admin)
+#   AIRFLOW_ADMIN_PASSWORD — Airflow admin password. REQUIRED: the api-server
+#                            refuses to start without it, rather than falling
+#                            back to a generated one that changes on restart.
 #   STAGING_DB_IP          — macvlan IP for staging Postgres (default 10.0.1.190)
 #   STAGING_FRONTEND_IP    — macvlan IP for staging frontend (default 10.0.1.151)
 
@@ -55,6 +64,12 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
+    volumes:
+      - unleash_cache:/app/.unleash
     depends_on:
       sc_database:
         condition: service_healthy
@@ -78,9 +93,20 @@ services:
       - FASTAPI_URL=http://backend:80/api
       - TYPESENSE_URL=http://typesense:8108
       - TYPESENSE_API_KEY=${TYPESENSE_SEARCH_KEY:-changeme}
+      # Payload CMS runs inside this container, in the `payload` schema of the
+      # staging database. Staging has its own stack, its own Postgres and its
+      # own admin account — never production's.
+      - DATABASE_URL=postgresql://${DB_USERNAME}:${DB_PASSWORD}@sc_database:5432/${DB_DATABASE_NAME}
+      - PAYLOAD_SECRET=${PAYLOAD_SECRET:?set PAYLOAD_SECRET in the staging Portainer stack environment}
+    volumes:
+      # Portainer prefixes volume names with the stack name, so this is
+      # automatically isolated from production's media.
+      - payload_media:/app/media
     depends_on:
       backend:
         condition: service_healthy
+      sc_database:
+        condition: service_healthy
     networks:
       backend: {}
       macvlan:
@@ -116,7 +142,23 @@ services:
   airflow-api-server:
     image: privaterepo.sitaru.org/tudor/school_compare-pipeline:staging
     container_name: sc_staging_airflow_api
-    command: airflow api-server --port 8080
+    # The simple auth manager generates a random password on first start and
+    # writes it to a file, so every container restart invalidates the last one.
+    # Writing the file ourselves from an environment variable makes the login
+    # deterministic. Airflow does not generate anything when the file exists.
+    #
+    # Built with python rather than echo/printf so a password containing quotes,
+    # backslashes or spaces is escaped correctly by json.dumps. An unset
+    # AIRFLOW_ADMIN_PASSWORD raises KeyError and the container exits: falling
+    # back to a generated password would silently undo the point of this.
+    command:
+      - bash
+      - -c
+      - |
+        set -euo pipefail
+        mkdir -p /opt/airflow
+        python -c "import json, os, pathlib; pathlib.Path('/opt/airflow/simple_auth_manager_passwords.json').write_text(json.dumps({os.environ.get('AIRFLOW_ADMIN_USER', 'admin'): os.environ['AIRFLOW_ADMIN_PASSWORD']}))"
+        exec airflow api-server --port 8080
     ports:
       - "8081:8080"
     environment:
@@ -128,6 +170,8 @@ services:
       AIRFLOW__API_AUTH__JWT_SECRET: "school-compare-staging-airflow-jwt-secret-key-long-enough-for-sha512"
       AIRFLOW__API_AUTH__JWT_ISSUER: airflow
       AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_USERS: "${AIRFLOW_ADMIN_USER:-admin}:admin"
+      AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_PASSWORDS_FILE: /opt/airflow/simple_auth_manager_passwords.json
+      AIRFLOW_ADMIN_PASSWORD: ${AIRFLOW_ADMIN_PASSWORD:?set AIRFLOW_ADMIN_PASSWORD in the Portainer stack environment}
       AIRFLOW__LOGGING__BASE_LOG_FOLDER: /opt/airflow/logs
       PG_HOST: sc_database
       PG_PORT: "5432"
@@ -212,3 +256,5 @@ volumes:
   postgres_data:
   typesense_data:
   airflow_logs:
+  unleash_cache:
+  payload_media:

docker-compose.portainer.unleash.yml

@@ -0,0 +1,73 @@
+# Portainer Stack Definition for School Compare — UNLEASH (feature flags)
+#
+# Deploy as a *separate* Portainer stack ("schoolcompare-unleash"), alongside
+# the production and staging stacks. It deliberately belongs to neither: a
+# staging redeploy must not be able to disturb production's flag state, and a
+# production redeploy must not disturb staging's.
+#
+# One instance serves both environments. Open-source Unleash ships with
+# `development` and `production` environments and environment-scoped client
+# tokens, so the same flag holds independent state in each — which is what
+# lets a feature be on in staging, where the E2E journeys exercise it, while
+# production stays dark.
+#
+# Portainer environment variables (set in Portainer UI -> Stack -> Environment):
+#   UNLEASH_DB_PASSWORD    — PostgreSQL password for the Unleash database
+#   UNLEASH_ADMIN_PASSWORD — initial admin password for the Unleash UI
+#   UNLEASH_IP             — macvlan IP for the Unleash server (default 10.0.1.152)
+
+services:
+
+  # ── PostgreSQL (Unleash's own; nothing else uses it) ──────────────────
+  unleash_db:
+    container_name: sc_unleash_postgres
+    image: postgres:16-alpine
+    environment:
+      POSTGRES_USER: unleash
+      POSTGRES_PASSWORD: ${UNLEASH_DB_PASSWORD}
+      POSTGRES_DB: unleash
+    volumes:
+      - unleash_postgres_data:/var/lib/postgresql/data
+    networks:
+      - unleash
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U unleash"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 10s
+    restart: unless-stopped
+
+  # ── Unleash server (UI + client API on 4242) ──────────────────────────
+  unleash:
+    container_name: sc_unleash
+    image: unleashorg/unleash-server:6
+    environment:
+      DATABASE_URL: postgres://unleash:${UNLEASH_DB_PASSWORD}@unleash_db:5432/unleash
+      DATABASE_SSL: "false"
+      INIT_ADMIN_API_TOKENS: ""
+      UNLEASH_DEFAULT_ADMIN_PASSWORD: ${UNLEASH_ADMIN_PASSWORD}
+    depends_on:
+      unleash_db:
+        condition: service_healthy
+    networks:
+      unleash: {}
+      macvlan:
+        ipv4_address: ${UNLEASH_IP:-10.0.1.152}
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:4242/health || exit 1"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 30s
+    restart: unless-stopped
+
+networks:
+  unleash:
+    driver: bridge
+  macvlan:
+    external:
+      name: macvlan
+
+volumes:
+  unleash_postgres_data:

docker-compose.portainer.yml

@@ -7,7 +7,15 @@
 #   ADMIN_API_KEY          — Backend admin API key
 #   TYPESENSE_API_KEY      — Typesense admin API key
 #   TYPESENSE_SEARCH_KEY   — Typesense search-only key (exposed to frontend)
-#   AIRFLOW_ADMIN_USER     — Airflow admin username (password auto-generated, see api-server logs)
+#   UNLEASH_URL            — http://<unleash-ip>:4242/api (empty = all flags off)
+#   UNLEASH_API_TOKEN      — Unleash *client* token, environment: production
+#   PAYLOAD_SECRET         — Payload CMS encryption secret. REQUIRED: long and
+#                            random. Changing it invalidates every admin
+#                            session. Staging MUST use a different value.
+#   AIRFLOW_ADMIN_USER     — Airflow admin username (default: admin)
+#   AIRFLOW_ADMIN_PASSWORD — Airflow admin password. REQUIRED: the api-server
+#                            refuses to start without it, rather than falling
+#                            back to a generated one that changes on restart.
 
 services:
 
@@ -44,6 +52,12 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
+    volumes:
+      - unleash_cache:/app/.unleash
     depends_on:
       sc_database:
         condition: service_healthy
@@ -67,9 +81,21 @@ services:
       - FASTAPI_URL=http://backend:80/api
       - TYPESENSE_URL=http://typesense:8108
       - TYPESENSE_API_KEY=${TYPESENSE_SEARCH_KEY:-changeme}
+      # Payload CMS runs inside this container. It reaches Postgres over the
+      # `backend` network and keeps its tables in the `payload` schema, so no
+      # pipeline operation on `public` can touch blog content.
+      - DATABASE_URL=postgresql://${DB_USERNAME}:${DB_PASSWORD}@sc_database:5432/${DB_DATABASE_NAME}
+      # Same :? form as AIRFLOW_ADMIN_PASSWORD: refuse to start rather than
+      # boot with an empty secret and silently accept forged sessions.
+      - PAYLOAD_SECRET=${PAYLOAD_SECRET:?set PAYLOAD_SECRET in the Portainer stack environment}
+    volumes:
+      # Blog images. Not reproducible from the pipeline — must be backed up.
+      - payload_media:/app/media
     depends_on:
       backend:
         condition: service_healthy
+      sc_database:
+        condition: service_healthy
     networks:
       backend: {}
       macvlan:
@@ -105,7 +131,23 @@ services:
   airflow-api-server:
     image: privaterepo.sitaru.org/tudor/school_compare-pipeline:prod
     container_name: schoolcompare_airflow_api
-    command: airflow api-server --port 8080
+    # The simple auth manager generates a random password on first start and
+    # writes it to a file, so every container restart invalidates the last one.
+    # Writing the file ourselves from an environment variable makes the login
+    # deterministic. Airflow does not generate anything when the file exists.
+    #
+    # Built with python rather than echo/printf so a password containing quotes,
+    # backslashes or spaces is escaped correctly by json.dumps. An unset
+    # AIRFLOW_ADMIN_PASSWORD raises KeyError and the container exits: falling
+    # back to a generated password would silently undo the point of this.
+    command:
+      - bash
+      - -c
+      - |
+        set -euo pipefail
+        mkdir -p /opt/airflow
+        python -c "import json, os, pathlib; pathlib.Path('/opt/airflow/simple_auth_manager_passwords.json').write_text(json.dumps({os.environ.get('AIRFLOW_ADMIN_USER', 'admin'): os.environ['AIRFLOW_ADMIN_PASSWORD']}))"
+        exec airflow api-server --port 8080
     ports:
       - "8080:8080"
     environment:
@@ -117,6 +159,8 @@ services:
       AIRFLOW__API_AUTH__JWT_SECRET: "school-compare-airflow-jwt-secret-key-long-enough-for-sha512"
       AIRFLOW__API_AUTH__JWT_ISSUER: airflow
       AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_USERS: "${AIRFLOW_ADMIN_USER:-admin}:admin"
+      AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_PASSWORDS_FILE: /opt/airflow/simple_auth_manager_passwords.json
+      AIRFLOW_ADMIN_PASSWORD: ${AIRFLOW_ADMIN_PASSWORD:?set AIRFLOW_ADMIN_PASSWORD in the Portainer stack environment}
       AIRFLOW__LOGGING__BASE_LOG_FOLDER: /opt/airflow/logs
       PG_HOST: sc_database
       PG_PORT: "5432"
@@ -201,3 +245,5 @@ volumes:
   postgres_data:
   typesense_data:
   airflow_logs:
+  unleash_cache:
+  payload_media:

docker-compose.yml

@@ -36,6 +36,10 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
     volumes:
       - ./data:/app/data:ro
     depends_on:
@@ -101,7 +105,23 @@ services:
   airflow-api-server:
     image: privaterepo.sitaru.org/tudor/school_compare-pipeline:latest
     container_name: schoolcompare_airflow_api
-    command: airflow api-server --port 8080
+    # The simple auth manager generates a random password on first start and
+    # writes it to a file, so every container restart invalidates the last one.
+    # Writing the file ourselves from an environment variable makes the login
+    # deterministic. Airflow does not generate anything when the file exists.
+    #
+    # Built with python rather than echo/printf so a password containing quotes,
+    # backslashes or spaces is escaped correctly by json.dumps. An unset
+    # AIRFLOW_ADMIN_PASSWORD raises KeyError and the container exits: falling
+    # back to a generated password would silently undo the point of this.
+    command:
+      - bash
+      - -c
+      - |
+        set -euo pipefail
+        mkdir -p /opt/airflow
+        python -c "import json, os, pathlib; pathlib.Path('/opt/airflow/simple_auth_manager_passwords.json').write_text(json.dumps({os.environ.get('AIRFLOW_ADMIN_USER', 'admin'): os.environ['AIRFLOW_ADMIN_PASSWORD']}))"
+        exec airflow api-server --port 8080
     ports:
       - "8080:8080"
     environment: &airflow-env
@@ -113,6 +133,8 @@ services:
       AIRFLOW__API_AUTH__JWT_SECRET: "school-compare-airflow-jwt-secret-key-long-enough-for-sha512"
       AIRFLOW__API_AUTH__JWT_ISSUER: airflow
       AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_USERS: "admin:admin"
+      AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_PASSWORDS_FILE: /opt/airflow/simple_auth_manager_passwords.json
+      AIRFLOW_ADMIN_PASSWORD: ${AIRFLOW_ADMIN_PASSWORD:-admin}
       PG_HOST: db
       PG_PORT: "5432"
       PG_USER: schoolcompare

docs/DEPLOY.md

@@ -98,6 +98,12 @@ fail the E2E gate. That's the point: staging absorbs the risk.
    pr-checks status checks (frontend, backend, builds, ai-review) to pass.
 5. **Bootstrap staging data via Airflow** (no prod dump — staging populates
    itself from source, exercising the pipeline image end-to-end):
+   - Set `AIRFLOW_ADMIN_PASSWORD` in the stack environment first. The
+     api-server refuses to start without it. Airflow's simple auth manager
+     otherwise generates a password on first start and writes it to a file, so
+     the login changes every time the container restarts; the stack writes that
+     file itself from this variable instead. `AIRFLOW_ADMIN_USER` defaults to
+     `admin`.
    - Open the staging Airflow UI (`http://<host>:8081`) and trigger, in order:
      `school_data_daily`, `school_data_monthly_ofsted`, then the manual-schedule
      `school_data_annual_ees` and `school_data_annual_idaci`.
@@ -150,3 +156,103 @@ token Gitea Actions provides automatically (`secrets.GITEA_TOKEN` — no setup
 needed), and fails the check only when a finding is rated
 **severe** (would break prod, leak data, or corrupt data). Minor findings are
 informational and never block a merge.
+
+## Rate limiting, and the Cloudflare gap
+
+Two independent limits protect the API:
+
+- **Per client**, via slowapi, keyed on `CF-Connecting-IP` (falling back to
+  `X-Forwarded-For`, then the peer address). 60/minute by default;
+  `/api/suggest` gets 120/minute because typing is bursty.
+- **Globally**, via `GlobalRateLimitMiddleware`: a fixed 60-second window over
+  all `/api/` traffic, `GLOBAL_RATE_LIMIT_PER_MINUTE` (default 3000),
+  independent of any client identity. Requests from `127.0.0.1` are exempt so
+  the container healthcheck cannot be starved into a restart loop.
+
+### Open: the origin must only accept Cloudflare
+
+`CF-Connecting-IP` is only meaningful for requests that actually reached the
+origin through Cloudflare, and **the application cannot verify that they did**.
+Anything able to reach the origin directly can set that header freely and, by
+rotating it, mint a fresh rate-limit bucket per request — defeating per-client
+limits on every endpoint.
+
+The global ceiling bounds the damage to total origin capacity. It does not fix
+the underlying gap, and nothing in the code can. Closing it needs one of:
+
+- **Authenticated Origin Pulls** — Cloudflare presents a client certificate the
+  origin requires, so non-Cloudflare traffic is refused at TLS.
+- **An origin firewall** restricted to Cloudflare's published IP ranges.
+
+Until one is in place, treat per-client limits as protection against accidents
+and ordinary load, not against a determined caller.
+
+## Feature flags (Unleash)
+
+Flag state lives in a self-hosted Unleash instance, deployed as its own
+Portainer stack from `docker-compose.portainer.unleash.yml`. It is separate
+from the application stacks on purpose — redeploying staging must not be able
+to disturb production's flags.
+
+The flags themselves are declared in `backend/flags.py`. Unleash holds the
+state; the registry holds the list. A flag in the UI that is not in the
+registry is orphaned and nothing reads it.
+
+### First-time setup
+
+1. Deploy the stack in Portainer. Set `UNLEASH_DB_PASSWORD`,
+   `UNLEASH_ADMIN_PASSWORD` and (optionally) `UNLEASH_IP`.
+2. Log in to the UI at `http://<UNLEASH_IP>:4242` as `admin`.
+3. Create one **client** API token per environment:
+   - `schoolcompare-staging`, environment **development**
+   - `schoolcompare-prod`, environment **production**
+
+   Client tokens, not admin tokens — the backend only reads.
+4. Put each token in the matching Portainer stack's `UNLEASH_API_TOKEN`
+   variable, and set `UNLEASH_URL` to `http://<UNLEASH_IP>:4242/api`.
+5. Redeploy the application stacks.
+
+### Adding a flag to Unleash
+
+**Unleash does not create flags by itself.** The SDK reads definitions from the
+server and never registers anything, and metrics for a flag the server has
+never heard of are discarded. So a flag declared in `backend/flags.py` will be
+evaluated on every request, stay `False` forever, and never appear in the UI
+until someone creates it there by hand.
+
+For each flag in the registry, create one in Unleash with:
+
+- **Name** — character for character what `backend/flags.py` declares.
+  snake_case, no hyphens or spaces. A typo produces a flag that looks correct
+  in the UI and is read by nothing.
+- **Type** — Release. No strategies, constraints or variants: these are plain
+  on/off switches, by design.
+
+### Turning a feature on
+
+Toggle the flag in the environment matching the stack you mean: **development**
+for staging, **production** for prod. The token in each stack is scoped to one
+environment, so toggling the other one has no visible effect.
+
+The SDK refreshes every 15 seconds, so the API reflects the change almost at
+once; the pages follow on their own schedule, below.
+
+A flip reaches school pages within about five minutes and place pages within
+the hour. Next's ISR does the propagating — it revalidates a route at the
+*lowest* `revalidate` among that route's fetches, which is 300s for
+`/school/[slug]` and 3600s for the place pages. There is no webhook, and
+adding one would only be worth it if flips ever needed to be instant.
+
+### When Unleash is unreachable
+
+Every flag evaluates to `False` and the site serves as though nothing were
+switched on. That is deliberate — an unfinished feature staying hidden is the
+safe direction — but it means a *released* feature disappears if a backend
+container cold-starts with an empty cache while Unleash is down. The SDK's
+disk cache is on a named volume so restarts keep last-known state, and flags
+are removed from the code within 90 days (enforced by a test), which bounds
+how long any feature is exposed to this.
+
+If `UNLEASH_URL` is unset, every flag is `False` and no connection is
+attempted. That is the correct behaviour for local development and CI, and it
+means the test suites need no flag server.

docs/superpowers/specs/2026-08-23-feature-flags-design.md

@@ -0,0 +1,294 @@
+# Feature Flags — Design
+
+**Date:** 2026-08-23
+**Status:** approved for planning
+**First consumer:** the last-distance-offered feature (`admission_distance`)
+
+## Goal
+
+Let work merge to `main` and deploy to production without becoming visible,
+so that releasing a feature stops being the same event as deploying it.
+
+The site has no way to do this today. A feature is either on `main` and live,
+or it is on a branch. That forces long-lived branches for anything not ready,
+and it makes every promotion to production an all-or-nothing decision about
+everything queued behind it.
+
+This is a **ship-dark** capability, not a kill switch. Flags are expected to
+flip on the order of once a month, by a person, deliberately. Nothing here is
+designed for flipping something off in seconds under pressure, and nothing
+here does percentage rollouts, user targeting or A/B tests — the site has no
+user identity to target.
+
+## Decision: Unleash
+
+Flag state is held in a self-hosted [Unleash](https://www.getunleash.io/)
+instance (Apache-2.0), not in the repository.
+
+A lighter option was considered and rejected by the project owner: a typed
+registry in each runtime with environment-variable overrides set in the
+Portainer stack files, which would have needed no new container and kept flag
+state in git. The argument for Unleash is that it provides a UI and an audit
+log without a deploy, and that flags are expected to become an ongoing
+operational tool rather than an occasional one.
+
+Two consequences follow from choosing a service, and this design exists mostly
+to handle them:
+
+1. **Flag state lives outside the repository.** `main` is no longer the whole
+   truth about what is switched on. The registry in §2 exists to bound that.
+2. **A flag can change without a deploy**, so nothing else clears the caches
+   that a deploy would have cleared. §4 establishes how long a flip takes to
+   become visible, and why that is short enough to need no extra mechanism.
+
+Also considered: Flagsmith (heavier — Django, Postgres and Redis), GrowthBook
+(requires MongoDB), and Flipt v2 (the closest conceptual fit, git-native, but
+now under the Fair Core Licence — source-available, not OSI open source).
+
+## 1. Topology
+
+A third Portainer stack, `docker-compose.portainer.unleash.yml`, holding
+`unleashorg/unleash-server` and its own PostgreSQL 16. It is on the macvlan so
+both application stacks can reach it, and it belongs to neither of them — a
+staging redeploy must not be able to disturb production's flag state, and vice
+versa.
+
+One instance serves both environments. Open-source Unleash ships with
+`development` and `production` environments and environment-scoped client
+tokens, so the same flag holds independent state in each: staging's FastAPI
+carries a `development` token, production's carries a `production` one.
+
+That property is what makes ship-dark testable. A feature can be **on in
+staging and off in production** for as long as it takes, which means the `e2e/`
+journeys exercise it against staging while production stays unchanged.
+
+## 2. The registry
+
+Unleash supplies flag *state* and the toggle UI. It does not supply the list of
+flags. `backend/flags.py` declares every flag the code knows about:
+
+```python
+@dataclass(frozen=True)
+class Flag:
+    name: str            # identical in the registry, in Unleash, and in JSON
+    description: str     # one line: what turning this on reveals
+    added: date          # for the staleness test in §8
+```
+
+**Every flag defaults to `False`.** There is no per-flag default field, because
+a flag that defaults on is not a ship-dark flag — it is a kill switch, and this
+design does not offer one. A single unconditional default also means the
+fallback path has no branching to get wrong.
+
+Three reasons the registry is not optional:
+
+- The Unleash SDK evaluates an unknown flag to `False`. Without a registry that
+  is an *undeclared* false — indistinguishable from a typo in a flag name.
+- `/api/flags` needs a key set to return when Unleash is unreachable. It cannot
+  enumerate flags it has never heard of.
+- A flag present in the Unleash UI but absent from the registry is orphaned,
+  and should be visibly so rather than quietly authoritative.
+
+**Naming.** One string, used unchanged as the registry key, the Unleash flag
+name, and the JSON key in `/api/flags`. It is snake_case, matching the API's
+existing convention (`admission_distance`, `rwm_expected_pct`) and the mirrored
+types in `nextjs-app/lib/types.ts`. No case transformation anywhere, so there
+is no mapping layer to get wrong.
+
+## 3. Read paths
+
+### Backend
+
+`backend/flags.py` wraps `UnleashClient` behind `is_enabled(name: str) -> bool`.
+
+Fail-closed is the default rather than something added: the Python SDK
+evaluates every flag to `False` until it has synchronised with the server. An
+unfinished feature therefore stays hidden when Unleash is unreachable, which is
+the correct direction for ship-dark.
+
+The SDK's fcache directory is mounted on a named volume so a container restart
+during an Unleash outage keeps last-known state rather than reverting a
+released feature to dark. The registry default remains `False`, so the worst
+case is a feature disappearing, never one appearing.
+
+### Frontend
+
+`nextjs-app/lib/flags.ts` exposes `getFlags(): Promise<Flags>`, a single
+server-side fetch of `/api/flags` returning a typed record. Server components
+only — no flag value reaches the browser bundle, and `package.json` gains no
+Unleash dependency. The Unleash client library stays entirely inside the
+service that already owns every other piece of data the frontend renders.
+
+The cost, named plainly: a purely front-end flag must still be declared in a
+Python file. It is a flat data edit rather than programming, and the return is
+one list, so nobody has to ask which service knows about a given flag.
+
+### `/api/flags` must not be publicly reachable
+
+`nextjs-app/app/api/[...path]/route.ts` proxies **everything** under `/api/` to
+FastAPI. Left alone, `https://www.schoolcompare.co.uk/api/flags` would return
+`{"admission_distance": false, ...}` — publishing the name and state of every
+unreleased feature, which defeats the purpose of shipping dark.
+
+The proxy therefore gains a denylist, and `flags` is on it: a request for a
+denied path returns 404 rather than being forwarded. Next's own `getFlags()` is
+unaffected because it calls `FASTAPI_URL` directly across the Docker network
+and never transits the public proxy.
+
+This is a general hole rather than a flags-specific one — the proxy will
+forward any future internal endpoint too — so the denylist is written as a
+named constant with a comment saying what belongs on it.
+
+## 4. Propagation
+
+**Time-based revalidation is sufficient. There is no webhook.**
+
+An earlier draft of this section specified two Unleash webhooks and a
+`revalidateTag('flags')` purge, on the premise that pages cache for seven days.
+That premise was wrong, and checking it removed the most complex part of the
+design.
+
+Next uses the **lowest** `revalidate` among a route's fetches to set the
+revalidation frequency of the whole route — the segment-level
+`export const revalidate` does not override a lower value inside it. Measured
+against this codebase:
+
+| Page family | Segment | Lowest fetch | Effective |
+|---|---|---|---|
+| `/school/[slug]` | 604800 | `fetchSchoolDetails` at 300 | **5 minutes** |
+| `/schools/*` | 604800 | `fetchNationalAverages` at 3600 | **1 hour** |
+
+The Unleash SDK polls every 15 seconds, so a flip reaches school pages within
+about five minutes and place pages within the hour, unaided. Flags flip
+monthly, by hand, deliberately. That is fast enough.
+
+What this removes: two webhook integrations, a `/api/revalidate-flags` route, a
+shared-secret-in-a-query-string scheme, an idempotency requirement against
+duplicate and out-of-order delivery, and a rule that every fetch in
+`nextjs-app/lib/` carry a cache tag. None of it has to be built, maintained, or
+kept correct as new fetches are added.
+
+**If instant flips are ever wanted**, the webhook is the way to add them, and it
+is purely additive — nothing in this design has to change first.
+
+### Two constraints this leaves behind
+
+**Never flag content on a `force-static` page.** `app/admissions/page.tsx`
+declares `export const dynamic = 'force-static'`, so it is baked at build time
+and never revalidates. A flag gating anything on such a page would not take
+effect until the next deploy, silently. If a flag ever needs to reach one, that
+page must first move to ISR.
+
+**A route-family flag still needs the sitemap rebuilt.** The sitemap is held in
+memory and rebuilt only at startup or via `POST /api/admin/regenerate-sitemap`.
+No flag in scope touches the sitemap (§6), so this is deferred with the route
+case rather than solved now — but a route flag must not ship without it, or the
+sitemap will advertise URLs that `notFound()`.
+
+## 5. What "off" means, per surface
+
+| Surface | Off |
+|---|---|
+| Route | `notFound()`, **and** absent from the sitemap, **and** absent from nav |
+| UI element | Not rendered; surrounding page byte-identical to today |
+| API field | Key **absent**, not `null` |
+| API endpoint | 404, not 403 |
+
+The three parts of the route rule move together or not at all. Submitting URLs
+to Google that return 404 is the bug fixed in PR #124, and a flag is a new way
+to reintroduce it.
+
+An API field is withheld **at the source**, never rendered-but-hidden. The
+precedent is already set in this codebase by commit `c9a1892`: `/api/schools/`
+is public and unauthenticated, so leaving a withheld field in the payload hands
+the record to anyone who opens the network tab.
+
+## 6. First consumer: `admission_distance`
+
+The last-distance-offered feature is merged to `main` and live on staging.
+Production has never received it: `/api/schools/100010` on production carries
+no `admission_distance` key, and no Distance section renders.
+
+It needs **exactly one gate** — `backend/app.py:809`, where the field is
+attached to the school payload:
+
+```python
+"admission_distance": (
+    supplementary.get("admission_distance")
+    if flags.is_enabled("admission_distance") else None
+),
+```
+
+The frontend follows with no change. `DistanceSection` already returns `null`
+when `admission_distance?.distance_m == null`, and `PrimarySchoolSections`
+already conditions the admissions block on `(admissions || admissionDistance)`.
+The off-state is the commonest state on the site — only 57 local authorities
+publish cut-off distances at all — so it is well covered by construction.
+
+The flag does not touch the sitemap: school pages exist either way.
+
+Intended lifecycle: default off, so production receives the code dark on the
+next promotion; on in the `development` environment so staging keeps testing
+it; flipped on in `production` when the owner chooses.
+
+**This flag exercises two of the three surfaces** in §5 — API field and UI
+element. No route case ships with it. The route rule is specified but unproven
+until a route-shaped flag exists, and should be treated as such.
+
+## 7. Testing
+
+**Backend unit.** The registry is well-formed; an unknown flag evaluates
+`False`; `/api/flags` returns every declared flag with its default when the
+SDK is unreachable; `admission_distance` is absent from the school payload when
+the flag is off and present when on.
+
+**Frontend unit.** `getFlags()` returns declared defaults when `/api/flags`
+fails, rather than throwing and taking the page with it.
+
+**E2E.** Journeys read `/api/flags` and gate flag-dependent assertions on it,
+matching the `test.skip` shape the suite already uses.
+
+One trap to avoid, worth stating because the existing distance journeys walk
+straight into it: they already skip when no school has a published figure, so
+with the flag off they would skip silently and the suite would go green. The
+gate must be explicit — **if `/api/flags` reports `admission_distance` on, then
+a school with a cut-off must be found**, converting a silent skip into a real
+assertion.
+
+## 8. Lifecycle
+
+A flag is temporary scaffolding, and the failure mode of every flag system is
+accumulation.
+
+The registry records the date each flag was added, and a backend test fails any
+flag older than **90 days**. Removing a flag means deleting the registry entry,
+the branches that read it, and the flag in the Unleash UI.
+
+Unleash SDK usage metrics stay enabled, so the UI shows which flags are still
+being evaluated — the evidence needed to retire one safely.
+
+## 9. Risks
+
+**Production gains a homelab dependency.** If Unleash is unreachable when a
+production container cold-starts with an empty cache, every flag evaluates
+`False` and any feature currently switched on disappears. The fcache volume
+covers restarts; the 90-day lifecycle rule bounds how long any feature is
+exposed to this. It is a real regression risk and the reason flags must be
+retired rather than left on indefinitely.
+
+**Flag state is not in git.** `main` no longer tells you what production is
+showing. The registry lists what *can* be flagged; only the Unleash UI says
+what *is*. This is inherent to the choice of a service.
+
+**A large promotion backlog exists.** Production is running the
+pre-SEO-programme build — no place pages, and a sitemap still declaring the
+apex host. The first promotion after this work ships that entire backlog. The
+flag isolates the distance feature from it and nothing else.
+
+## Out of scope
+
+- Percentage rollouts, user targeting, A/B testing, and Unleash strategies
+  beyond simple on/off. Flags are booleans.
+- Pipeline and dbt flags. Airflow and dbt are not flag consumers.
+- Client-side flag evaluation. Flags are server-side only.
+- Automatic flag removal. The staleness test reports; a person deletes.

docs/superpowers/specs/2026-08-26-school-autosuggest-design.md

@@ -0,0 +1,294 @@
+# School Autosuggest — Design
+
+**Date:** 2026-08-26
+**Status:** approved for planning
+**Depends on:** the feature-flag layer (PR #125, merged)
+
+## Goal
+
+Suggest schools by name as someone types in the site's main search box, so a
+parent who knows the school they want reaches it in one step instead of
+searching, scanning a result list, and clicking.
+
+Scope is **schools only**. Places and postcodes were considered and excluded —
+see *Out of scope*.
+
+## The finding that shapes everything
+
+The site's rate limiter does not do what it looks like it does.
+
+`limiter = Limiter(key_func=get_remote_address)` with `60/minute` reads
+`request.client.host`. In staging and production the backend has no published
+ports and sits on the internal `backend` network, so its only caller is the
+Next proxy — and `request.client.host` is therefore **the Next container**, for
+every browser user on the site.
+
+Measured against staging: 70 concurrent requests to `/api/schools` returned
+**60 × 200 and 10 × 429**. One machine consumed the whole site's budget for
+that minute.
+
+Autosuggest is the worst possible feature to build on that. One person typing
+"st marys primary" produces six to eight debounced requests; **eight concurrent
+searchers would 429 the site.** The compare modal's search-as-you-type already
+shares this bucket, so the exposure exists today — autosuggest makes it
+certain.
+
+Fixing the keying is therefore part of this work, not a follow-up.
+
+## 1. Rate-limit keying
+
+Both environments sit behind Cloudflare (`server: cloudflare`, `cf-ray` present
+on staging and production). Cloudflare sets `CF-Connecting-IP` on every request
+to the origin and **overwrites any client-supplied value**, which makes it
+trustworthy in a way a parsed `X-Forwarded-For` chain is not.
+
+```python
+def client_key(request: Request) -> str:
+    """Rate-limit bucket: the real caller, not the proxy in front of them."""
+    cf = request.headers.get("cf-connecting-ip")
+    if cf:
+        return cf.strip()
+    xff = request.headers.get("x-forwarded-for")
+    if xff:
+        return xff.split(",")[0].strip()
+    return get_remote_address(request)
+```
+
+`nextjs-app/app/api/[...path]/route.ts` already forwards every inbound header
+except `host` and `connection`, so `CF-Connecting-IP` reaches the backend with
+no proxy change.
+
+**This header is trustworthy only for traffic that actually passed through
+Cloudflare, and nothing in the application can verify that it did.** An earlier
+draft of this section claimed Cloudflare "replaces the header, so a browser
+cannot forge it", and that only the `X-Forwarded-For` fallback was forgeable.
+That was wrong. Cloudflare does overwrite the header *on requests it handles* —
+but a caller reaching the origin directly sets whatever it likes, and this
+process cannot distinguish an edge-set header from an attacker-set one. Both
+headers are equally forgeable in that scenario.
+
+The consequence is sharper than a weakened defence. An attacker rotating
+`CF-Connecting-IP` per request mints a fresh rate-limit bucket every time and
+evades per-client limits entirely — including on the DataFrame-heavy
+`/api/schools`. Against abuse that is *worse* than the shared bucket it
+replaced, which at least capped everyone at 60/minute together.
+
+Two mitigations, and they are not interchangeable:
+
+1. **The real fix is at Cloudflare** — Authenticated Origin Pulls, or an origin
+   firewall that refuses connections not from Cloudflare's ranges. Only the
+   edge can vouch for its own header. This is infrastructure work and is not
+   part of this change; it is the thing that makes the header mean anything.
+2. **The ceiling in §1.1 bounds what evading the keying can achieve** while
+   that remains open. It does not make the header trustworthy — it makes
+   trusting it survivable.
+
+The backend being unreachable from outside the Docker network is a real second
+layer, but it depends on the ingress path in front of the frontend, which this
+design does not control and should not assume.
+
+### 1.1 The ceiling, which is back
+
+The shared bucket was acting as an accidental global throttle on a
+single-process uvicorn backend that filters a 25,000-row DataFrame in-process.
+Correct per-user keying removes it: the origin becomes reachable at 60/min *per
+user* rather than 60/min in total, and — per above — at an unbounded rate by
+anyone willing to rotate a header.
+
+An earlier draft dropped the in-app ceiling, arguing it belonged at Cloudflare.
+That argument assumed the keying was sound. It is not, so the ceiling is
+load-bearing rather than redundant, and it ships here:
+
+`GlobalRateLimitMiddleware` counts all `/api/` requests in a fixed 60-second
+window against `global_rate_limit_per_minute` (3000), independent of any client
+identity, and refuses with a 429 that names capacity rather than the client —
+an operator has to be able to tell "one noisy client" from "the origin is
+saturated". It is registered last so it is outermost: a ceiling that applies
+after the expensive work has run is not a ceiling.
+
+slowapi cannot express this. `default_limits` and `application_limits` are both
+evaluated with the same `key_func`, making them per-client rather than global,
+and `application_limits` only apply with `SlowAPIMiddleware` installed, which
+this app does not use. Hence the explicit middleware — about thirty lines, and
+obviously correct, which is what a backstop needs to be.
+
+Requests from `127.0.0.1` are exempt. The container healthcheck runs
+`curl http://localhost:80/api/data-info` from inside the container, and
+starving it would fail the check, restart the container, and turn a load spike
+into an outage loop. The exemption keys on the peer address, never the `Host`
+header, which the caller sets.
+
+3000/minute is an estimate, not a measurement, and worth revisiting against
+real traffic.
+
+### Per-user limits
+
+Per-user fairness and origin protection are different jobs, and this design now
+does both separately: the ceiling above for the origin, and per-route limits
+for fairness. Conflating them is what produced the original behaviour, where
+one bucket served the whole internet.
+
+The existing 60/minute default is unchanged, and `/api/suggest` gets
+120/minute. Both are estimates rather than measurements, and are a starting
+point to revisit once the keying is correct enough for real per-user traffic to
+be visible — which it was not before, because everyone shared one bucket.
+
+## 2. `GET /api/suggest`
+
+A dedicated endpoint, not a mode of `/api/schools`.
+
+The existing search path calls Typesense for URNs and then filters, ranks and
+sorts the full in-memory DataFrame — a pandas pass per keystroke, holding the
+GIL and blocking other requests in the same worker. Suggestions need none of
+it: `urn`, `school_name`, `phase`, `school_type`, `local_authority`,
+`postcode` and `ofsted_rating` are all already in the Typesense document
+(`pipeline/scripts/sync_typesense.py`).
+
+```
+GET /api/suggest?q=<query>&limit=8
+→ 200 {"suggestions": [
+      {"urn": 100010, "school_name": "Brecknock Primary School",
+       "local_authority": "Camden", "postcode": "NW1 1AA",
+       "phase": "Primary", "school_type": "Community school"}
+  ]}
+```
+
+- **Under two characters** returns `{"suggestions": []}` with 200. The
+  keystroke path never returns an error for ordinary input.
+- **Typesense unavailable** returns `{"suggestions": []}` with 200. There is
+  deliberately **no DataFrame fallback**: the substring scan `/api/schools`
+  falls back to is precisely the cost this endpoint exists to avoid, and a
+  silent 25,000-row scan per keystroke is worse than no suggestions.
+- **`limit` is clamped** to 20. It is a public endpoint.
+- **Rate limit `120/minute`** per client, not the default 60. A 200 ms
+  debounce tops out near 5 requests/second while someone is actively typing,
+  but averages far below that across a real search; 120 leaves headroom for
+  bursts while still bounding one client.
+- **Local authority is part of the payload, not decoration.** There are many
+  schools called "St Mary's"; a suggestion list without the authority is
+  unusable for exactly the queries autosuggest is meant to serve.
+
+### Caching
+
+`CACHE_RULES` gains `("/api/suggest", (60, 3600, 86400))`. Prefix queries
+repeat enormously across users and school names change once a year.
+
+The client fetch must **not** use `cache: "no-store"`. The compare modal does,
+and copying that pattern would throw away both the browser cache and the ETag
+304s the existing `CacheAndETagMiddleware` already provides.
+
+Both environments currently report `cf-cache-status: DYNAMIC` — Cloudflare
+ignores the `Cache-Control` the API already sends, because it does not cache
+dynamic paths by default. **A Cloudflare Cache Rule for `/api/suggest*` would
+let the edge absorb most of this traffic and never reach the origin.** That is
+a dashboard change, it is optional, and nothing here depends on it.
+
+## 3. The combobox
+
+This is an ARIA combobox, not a text input with a list underneath.
+
+**Files.** `FilterBar.tsx` is already long. The work splits three ways:
+`hooks/useSchoolSuggest.ts` owns fetching, debouncing and cancellation;
+`components/SuggestList.tsx` owns rendering and ARIA; `FilterBar.tsx` wires
+them to the existing input and form.
+
+**Fetching.** 200 ms debounce; minimum two characters; an `AbortController`
+cancels the superseded request on every keystroke. Cancellation is not an
+optimisation — without it, a slow response for `"st"` can land after the fast
+one for `"st marys"` and replace a correct list with a stale one.
+
+**Suppressed during postcode entry.** The box takes a school name *or* a
+postcode, and `isValidPostcode` already distinguishes them. Suggestions do not
+appear once the value parses as a postcode.
+
+**Keyboard.** `ArrowDown`/`ArrowUp` move the active option, `Escape` closes and
+keeps the typed text, `Tab` closes. `Enter` **with an option active** navigates
+to that school's page. `Enter` **with none active** submits the free-text
+search exactly as it does today — the existing behaviour is preserved, not
+replaced.
+
+**ARIA.** `role="combobox"` with `aria-expanded` and `aria-controls` on the
+input, `aria-activedescendant` pointing at the active option, `role="listbox"`
+on the list and `role="option"` on each row.
+
+**Both instances get it.** `HomeView` renders `FilterBar` twice — hero and
+sticky — from one component, so there is one implementation.
+
+## 4. Behind a flag
+
+Flag `school_autosuggest`, declared in `backend/flags.py`, default off.
+
+This is the most-used control on the site and the first change to it in a
+while. `app/page.tsx` is an async server component, so it reads the flag and
+threads it to `FilterBar` through `HomeView` — two prop hops, explicit, no
+client-side flag read.
+
+Off means the input behaves exactly as it does today: no listener, no fetch, no
+markup. Not a rendered-then-hidden dropdown.
+
+The rate-limit keying is **not** flagged. It is a correctness fix that should
+apply whether or not autosuggest is on, and flagging it would mean shipping a
+known-wrong limiter into production deliberately.
+
+## 5. Analytics
+
+`search_submitted` already carries `via: 'input'`. Selecting a suggestion fires
+it with `via: 'suggestion'` plus the chosen `urn`, so the obvious question —
+does this actually help, or do people ignore it — has an answer in the data
+rather than an opinion.
+
+## 6. Testing
+
+**Backend.** `client_key` prefers `CF-Connecting-IP`, falls back through
+`X-Forwarded-For` to the remote address, and two different values get two
+different buckets. `/api/suggest` returns matches, returns empty below two
+characters, returns empty and 200 when Typesense is unavailable, and clamps
+`limit`. That it never touches the DataFrame is asserted by making
+`load_school_data` raise and requiring the endpoint to answer anyway.
+
+**Frontend.** The hook debounces, aborts superseded requests, and drops a
+late-arriving response for a stale query. The list renders the ARIA
+attributes. Keyboard navigation moves the active option; `Enter` on an option
+navigates; `Enter` on none submits the search.
+
+**E2E.** With the flag on, typing a known school name shows it and selecting it
+lands on that school's page. With the flag off, no combobox markup exists.
+Gated on the flag the same way the distance journeys are — read the observable
+effect, since `/api/flags` is denied to the public.
+
+## 7. Risks
+
+**Removing the accidental throttle.** Covered in §1. Correct per-user keying
+means the origin is reachable at 60/minute *per user* where it was 60/minute
+in total, and no in-app global cap replaces it — that job goes to Cloudflare,
+which is not done as part of this change. Until it is, a determined caller
+with many source addresses can put more load on a single-process origin than
+they can today. Against this site's traffic that is a theoretical risk rather
+than a live one, but it is a real one and it is the price of the fix.
+
+**Cloudflare bypass — the open one.** If the origin is reachable without
+passing through Cloudflare, `CF-Connecting-IP` is attacker-controlled, and
+rotating it per request defeats per-client limits on every endpoint. The
+ceiling in §1.1 bounds the damage to the origin's total capacity; it does not
+restore per-client fairness under attack, and it cannot. Closing this properly
+means Authenticated Origin Pulls or an origin firewall restricted to
+Cloudflare's published ranges — infrastructure work, outside this change, and
+the single most valuable follow-up here.
+
+**Typesense becomes user-visible.** Today a Typesense outage degrades search to
+a slow substring match. With autosuggest it also means the dropdown silently
+stops appearing. That is the correct failure — quiet, not broken — but it makes
+Typesense health worth monitoring in a way it was not before.
+
+## Out of scope
+
+- **Place suggestions.** The 2,646 town, authority and outcode pages are a
+  strong candidate and would route people onto the pages W2 built, but they
+  live in the place registry rather than Typesense, so it is a second index and
+  a ranking rule for comparing two kinds of result. Worth its own change.
+- **Postcode completion.** Would put postcodes.io in the keystroke path, with
+  its own latency and rate limits.
+- **The compare modal.** It already has search-as-you-type. Converting it to
+  this component is a reasonable follow-up, not part of this.
+- **Recent or popular searches.** No storage for either, and no evidence yet
+  that they are wanted.

docs/superpowers/specs/2026-08-28-destination-measures-design.md

@@ -0,0 +1,399 @@
+# Destination Measures — Design
+
+**Date:** 2026-08-28
+**Status:** awaiting review
+**Scope:** secondary school detail pages only
+
+## Goal
+
+Say what happened to a school's leavers after they left. Two sections on the
+secondary template:
+
+- **After Year 11** — every secondary, from the KS4 destination measures
+- **After the sixth form** — sixth-form schools only, from the 16-18 measures
+
+This replaces the "Post-16 destination data coming soon" placeholder standing in
+`nextjs-app/components/school/SecondaryAdmissionsSection.tsx:117` since the exam
+phase taxonomy work, and fills the `ks5_destinations_pct` slot specified but
+never built in `2026-07-07-exam-phase-taxonomy-design.md:201`.
+
+Mockup, with all three data states live:
+<https://claude.ai/code/artifact/5be149d6-252f-473c-9a4f-4c36b05161b0>
+
+## The finding that shapes everything
+
+**Suppression is per cell, and the cells sum to the cohort.**
+
+DfE withholds a figure it considers disclosive by writing `c`. It does this at
+the level of an individual destination category, not the whole school, and it
+publishes the cohort total alongside. The categories form a clean partition. So
+where exactly one category is suppressed, subtracting the published ones from the
+cohort recovers it exactly.
+
+Verified against three real schools in the 2022/23 file:
+
+| School | URN | Withheld | Recovers to |
+|---|---|---|---|
+| North East Futures UTC | 145900 | School sixth form | **3 pupils** |
+| Whitley Bay High School | 108638 | Further education | **18 pupils** |
+| St Matthew's RC High School | 148389 | School sixth form | **4 pupils** |
+
+Those are the precise numbers the `c` exists to hide, and in a random 400-school
+sample **22% of mainstream secondaries** have exactly one suppressed category in
+their disadvantaged group. This is the normal case, not an edge case.
+
+Three rules follow, and everything else in this document is downstream of them.
+
+**R1 — Never *publish* enough to derive a remainder.**
+
+An earlier draft of this rule said "never *render* a derived remainder", and
+that was the defect code review caught in PR #137. Not drawing a number does
+nothing to stop it being computed: `GET /api/schools/{urn}` is public and
+unauthenticated, so anything in the payload is published whatever the UI
+chooses to draw. The rendering guards shipped; the payload still carried the
+cohort and every published category, and `cohort - sum(published)` returned
+Whitley Bay's withheld figure exactly.
+
+The rule is therefore about the serialiser, and the UI guards are a second line
+of defence behind it. Two identities have to be closed:
+
+- within a pupil group the categories sum to the cohort, so a group with
+  exactly **one** suppressed category gives it away;
+- across groups, disadvantaged + other = all for every category, so a category
+  suppressed in exactly **one** of the three gives itself away.
+
+`_mask_for_disclosure` applies DfE's own answer — secondary suppression —
+withholding a companion cell until every row and every column hides either none
+or at least two. It iterates, because each new suppression can break the other
+identity, and terminates because cells are only ever added.
+
+The companion must carry pupils. Suppressing a zero looks like secondary
+suppression and protects nothing: the residual still equals the original
+withheld figure.
+
+Where no companion can do the job — a sparse cohort whose every other category
+is `not_applicable`, routine in special schools and alternative provision — the
+pupil group is **dropped from the payload entirely**. A first version simply
+returned at that point with the violation intact and no signal, which review
+caught: a disclosure-control pass that fails silently is worse than none,
+because everything downstream trusts it. The function now cannot terminate
+except in a state where `disclosure_invariant_holds()` is true, and an
+exhaustive test sweeps all 81 suppression patterns of a four-category group to
+prove it.
+
+Measured cost on the 400-school sample: the all-pupils bar survives on **94%**
+of mainstream secondaries rather than 100%. That is the price of not
+republishing what DfE withheld.
+
+**R2 — Never aggregate across a suppression boundary.** Summing published
+components to fill a gap is R1 with extra steps.
+
+DfE's own aggregates (`Sustained education destination`, `Sustained education,
+employment & apprenticeships`) are ingested but **not served**. An aggregate
+spanning exactly one suppressed component names it, and nothing renders them
+today — an unused field that leaks is not a trade-off worth carrying. They can
+be re-added with their own guard if the fallback ladder is ever built.
+
+**R3 — The three pupil groups are one disclosure surface, not three.**
+Disadvantaged and Not-known-to-be-disadvantaged partition All pupils, so
+rendering any *two* of them recovers the third. Where a category is suppressed in
+the disadvantaged group, it must therefore also be withheld from **all other
+pupils** — the all-pupils view is the primary one and keeps it.
+
+This costs almost nothing, because DfE already applies the same masking: across
+the sample, 493 of 498 suppressed disadvantaged cells were suppressed in the
+other group too. The mart enforces the remaining 5, which fell on 2 schools of
+262. **The all-pupils bar is unaffected** — masking the whole page wherever the
+disadvantaged group is thin would remove the bar from 80% of schools, and is not
+what this rule says.
+
+R1 and R2 both hold within a group and still leak across the switch, which is why
+R3 is stated separately.
+
+### The convention that would break this quietly
+
+`macros/safe_numeric.sql` coerces every EES sentinel — `z`, `c`, `x`, `q`, `u` —
+to `NULL`, deliberately and correctly for attainment, where "suppressed" and "no
+data" are equally unrenderable. Here they are not the same thing: one must print
+*withheld*, the other must print nothing at all, and the difference is what keeps
+R1 enforceable.
+
+**`safe_numeric` must not be used on destination counts.** The staging model
+keeps the sentinel in a companion status column. This is the single most likely
+way for this feature to regress into a disclosure, so it gets its own dbt test.
+
+## What is actually available
+
+Measured against the EES public API (open, no key). Both datasets carry
+`geographicLevel: School` with `urn` on every location option, so the join to
+`dim_school` is direct.
+
+| | KS4 | 16-18 |
+|---|---|---|
+| Dataset id | `019d4f41-22d1-71b2-a1a7-f3b91026815b` | `019d4e73-6440-7523-b60c-bfab1ad4a30d` |
+| Rows | 1,871,739 | 3,862,658 |
+| Institutions | 4,946 | 3,065 |
+| Time periods | 2009/10–2022/23 | 2016/17–2022/23 |
+
+**Destination categories (KS4).** School sixth form · Sixth form college ·
+Further education · Other education destination · Sustained apprenticeships (with
+level breakdown) · Sustained employment destination · Not recorded as a sustained
+destination · Activity not captured. Plus the aggregates `Sustained education
+destination` and `Sustained education, employment & apprenticeships`.
+
+**16-18 adds** UK higher education institution and FE split by level, which is
+what makes the post-16 section worth having.
+
+**Breakdowns.** `Disadvantage Status` gives Disadvantaged / Not known to be
+disadvantaged / Total — exactly the three-way switch. Sex, ethnicity, FSM status,
+prior attainment and SEN provision also travel in the same table; we ingest none
+of them.
+
+**Indicators.** Both counts and percentages, plus the cohort size. Bar widths use
+the counts — the published percentages do not sum to 100.
+
+### Coverage, and what degrades
+
+Random 400-school sample, 2022/23, mainstream secondaries (n=262):
+
+| View | As published by DfE | After R1–R3 masking | Consequence |
+|---|---|---|---|
+| All pupils, all categories | 100% | **94%** | Bar works nearly everywhere |
+| Disadvantaged, headline rate | 95% | 95% | Gap panel works |
+| Disadvantaged, three grouped cards | 68% | 68% | Degrades card by card |
+| Disadvantaged, all six categories | 20% | **20%** | Bar unusable for this group |
+
+The middle column is what the site actually serves. Masking costs the
+all-pupils bar on 6% of mainstream secondaries — those are schools where a
+category was suppressed in exactly one pupil group and no non-zero companion
+existed below the all-pupils row.
+
+Special schools and alternative provision are far worse: 13% and 41% respectively
+have the whole cohort suppressed even for all pupils. The empty state is
+load-bearing, not defensive.
+
+## The display
+
+Question-led. Three cards over one bar, with the cards acting as a lens on the
+bar rather than a summary beside it — hovering a card dims the bar, table and
+England reference to the categories that card is built from. The full mockup is
+linked above; what matters for implementation:
+
+**The headline is not the sustained rate.** That figure sits between 92% and 97%
+for nearly every school in England. The mix is what varies, so the mix leads.
+
+**The grouping is ours, not DfE's.** "Academic route" = school sixth form +
+sixth-form college; "College" = FE and other colleges; "Work" = apprenticeship +
+employment. This is the most arguable thing on the page, so it lives in one place
+in `lib/destinations.ts`, is explained in a tooltip, and is reversible in one
+edit.
+
+**The absence is hatched neutral, never a colour.** "Activity not captured" means
+no record in the sources DfE holds — it includes independent schools, moving
+abroad and private training. Colouring it as a bad outcome would be a factual
+error rendered in CSS. The hatch also fixes a real contrast problem: neutral
+against the employment blue failed CVD separation at ΔE 7.6, and texture is the
+secondary encoding that rescues it. Every other adjacent pair clears ΔE 10.9
+under protanopia.
+
+**Colour tokens.** Education is one hue in three steps (school-like to
+college-like); apprenticeship and employment are separate hues. Six new tokens in
+`globals.css`, defined in both themes, per the existing token discipline.
+
+**The disadvantage split rides the same control.** One visualisation serving
+three cohorts, with the England reference repointing to the matching national
+group. The gap statement stays visible below the bar whatever is selected,
+because a gap nobody clicks on is a gap nobody sees.
+
+## Data model
+
+### Extraction
+
+A new `tap-uk-ees-destinations` extractor, separate from `tap-uk-ees`. The
+existing tap downloads a release ZIP and reads a CSV inside it; the destinations
+files are far larger than we need and the query API filters server-side, so this
+one POSTs to `/v1/data-sets/{id}/query` and pages through results.
+
+With every dimension pinned — destination measures, disadvantage status, sex
+Total, characteristic topic Total — one year returns **252,610 rows** across all
+geographic levels. Three school-level years is comfortably tractable.
+
+Pinning is mandatory, not an optimisation: leaving the characteristic dimensions
+unconstrained returned 45 rows where 9 were wanted, because every breakdown
+shares one table.
+
+The tap emits the raw value as text. **It does not coerce `c`.**
+
+### Staging
+
+`stg_ees_ks4_destinations` / `stg_ees_ks5_destinations`. Each raw value becomes
+two columns:
+
+```sql
+case when raw ~ '^-?[0-9]+(\.[0-9]+)?$' then raw::numeric end as pupils,
+case
+    when raw ~ '^-?[0-9]+(\.[0-9]+)?$' then 'published'
+    when lower(trim(raw)) = 'c'        then 'suppressed'
+    else 'not_applicable'
+end as status
+```
+
+### Marts
+
+`fact_ks4_destinations` and `fact_ks5_destinations`, **long format**:
+
+```
+urn, year, pupil_group, destination_category, cohort_pupils, pupils, percentage, status
+```
+
+This departs from the wide house pattern (`fact_ks4_performance` and friends) on
+purpose. `pupil_group` is a genuine third dimension; going wide would need three
+sets of every column, and R2 is far easier to test on rows than on columns.
+
+Roughly 8 categories × 3 groups × 4,946 schools × 3 years ≈ 356k rows.
+
+`fact_destination_national` carries the same grain for England, so the page's
+England reference repoints with the switch.
+
+### dbt tests
+
+- `assert_destinations_no_derived_remainder` — for every (urn, year,
+  pupil_group) with exactly one suppressed category, assert no aggregate row
+  exists that would let the residual be recovered. **This is the R1 guard.**
+- `assert_destinations_group_masking` — for every (urn, year, category), if the
+  disadvantaged group carries `suppressed`, so does the other-pupils group.
+  **This is the R3 guard**, applied in the mart so no consumer can reach an
+  unmasked combination.
+- `assert_destination_status_null_agreement` — `pupils is null` wherever
+  `status != 'published'`, and never null where it is.
+- `assert_destinations_join_dim_school` — no orphaned URNs, matching the
+  existing `assert_no_orphaned_facts` pattern.
+
+## API
+
+`GET /api/schools/{urn}` gains a `destinations` block:
+
+```json
+{
+  "ks4": {
+    "cohort_year": "2022/23",
+    "published": "2026-04",
+    "groups": {
+      "all":           { "cohort": 180, "categories": [ … ], "aggregates": { … } },
+      "disadvantaged": { … },
+      "other":         { … }
+    }
+  },
+  "ks5": { … }
+}
+```
+
+Each category carries `pupils`, `percentage` and `status`. **The serialiser never
+emits a computed remainder**, and a backend test asserts that a group containing a
+suppressed category serialises no total that closes the gap.
+
+`null` for the whole block where nothing is published — the frontend renders the
+empty state from its absence, not from a sentinel.
+
+## Frontend
+
+| File | Kind | Job |
+|---|---|---|
+| `lib/destinations.ts` | pure | Category list, the academic/college/work grouping, `canAggregate()` enforcing R2, percentage derivation from counts |
+| `components/school/DestinationsSection.tsx` | server | Section shell, renders **all pupils** into the HTML |
+| `components/school/DestinationsView.tsx` | client | Cohort switch, card↔bar linkage |
+| `components/school/Post16DestinationsSection.tsx` | server | Year 13 section, sixth-form schools only |
+| `app/globals.css` | tokens | Six destination colours, both themes |
+
+Server-first matches the directory's existing discipline — every component in
+`components/school/` is a server component except `AdmissionsViewToggle`, which
+is the precedent this follows. All-pupils figures are in the HTML for crawlers
+and for no-JS; only the switch and the hover linkage need the client.
+
+`lib/schoolSections.ts` gains `hasKs4Destinations` / `hasKs5Destinations` flags
+and the nav items, following the existing `computeSchoolFlags` pattern.
+
+**Placement** on the secondary template: GCSE results → After Year 11 → After the
+sixth form → admissions. Destinations follow attainment because they answer "and
+then what happened".
+
+**Dating.** The latest destination year is 2022/23, published April 2026, while
+the site's newest KS4 year is 2024/25. The section header states its own cohort
+year, or it reads as stale data next to the GCSE section above it.
+
+## Edge states
+
+| State | Frequency | Behaviour |
+|---|---|---|
+| Whole cohort suppressed | 13% of special, 41% of AP | Section renders the explanation, no chart |
+| Some categories withheld | 80% of disadvantaged views | Cards degrade individually; **no bar**; table marks withheld rows |
+| Disadvantaged group suppressed entirely | 5% | Switch drops to two options, gap panel not rendered |
+| No sixth form | — | Post-16 section not rendered at all — absence is correct, a "no data" placeholder would imply something is missing |
+| School too new | — | "First figures expected in 2026", not a bare no |
+
+## Testing
+
+Per CLAUDE.md, user-facing behaviour extends `e2e/` in the same PR.
+
+**Unit** — `lib/destinations.ts` is where R1 and R2 live, so it carries the
+heaviest tests: `canAggregate()` refuses a group containing one suppressed cell,
+allows one spanning two, and the bar builder refuses to emit segments for any
+group with suppression. These are the tests that must fail loudly if someone
+later "fixes" a gap in the chart.
+
+**dbt** — the three tests above.
+
+**Backend** — the serialiser emits no closing total for a partially suppressed
+group.
+
+**E2E** — a school with full data renders three cards and a bar; a school with a
+partially suppressed disadvantaged group renders the withheld state and **no bar
+element**; a suppressed school renders the explanation; a school with no sixth
+form renders no post-16 section.
+
+Note the staging caveat: mart changes are inert until the Airflow pipeline runs,
+and the staging E2E gate runs post-merge.
+
+## Out of scope
+
+- **Compare view and rankings.** The long mart shape supports both; neither is
+  built here. Flagged because "% to a school sixth form" is a plausible rankings
+  metric and the mart shape should not have to change to allow it.
+- **Ethnicity, sex, SEN and prior-attainment breakdowns.** Available in the same
+  file, ingested deliberately not at all — each is a separate editorial decision
+  about what a school page should assert.
+- **Longer term destinations** (3 and 5 years out) and **Progression to higher
+  education** — separate publications, worth a later look for sixth forms.
+- **Primary schools.** No KS2 destination measures publication exists; DfE
+  tracking starts at KS4. Naming the secondaries a primary's leavers go to needs
+  the National Pupil Database, which is not publishable at that grain.
+
+## Risks
+
+**A later change reintroduces the disclosure.** The likeliest routes are
+applying `safe_numeric` to a destination column for consistency, adding a
+`coalesce` in a mart, or — as happened in review — enforcing a disclosure rule
+at the rendering layer instead of the publishing layer. Mitigation is the dbt
+tests plus `backend/tests/test_destinations_api.py`, which reconstructs the
+residual the way an attacker would and asserts it no longer resolves.
+
+**The two-year lag reads as staleness.** Mitigated by dating the cohort in the
+section header rather than only in a tooltip.
+
+**Sixth-form retention will be misread.** "41% went to a school sixth form" says
+nothing about *which* school. The published file reports destination type, never
+destination institution. Copy must never imply "stayed on here", and the tooltip
+should say so.
+
+**Section length.** The secondary template is already long and this adds two
+sections. If it becomes a problem the post-16 section is the one to collapse
+behind a disclosure, not the Year 11 one.
+
+## Open questions
+
+1. Is the disadvantage split its own section or a sub-block inside the
+   destinations section? Modelled as a sub-block; it is the most differentiating
+   figure on the page and the most easily misread on a small cohort.
+2. Do we ingest the apprenticeship level breakdown (intermediate / advanced /
+   higher) now, or collapse to one apprenticeship figure and revisit? Collapsed
+   in this design.

docs/superpowers/specs/2026-09-02-about-and-blog-design.md

@@ -0,0 +1,368 @@
+# Giving schoolcompare a human author: an About page and a blog
+
+**Date:** 2026-09-02
+**Status:** Design — awaiting review
+**Scope:** A named author for the site, an `/about` page, and a Payload-CMS-backed
+blog at `/blog`.
+
+## Why
+
+The site reads as synthetic. Not because of its tone, but because of three
+specific absences:
+
+1. **Nobody is accountable for the numbers.** There is no author, no statement
+   of why the site exists, and no one who can be wrong. The only human trace on
+   the entire site is `contact@schoolcompare.co.uk` in the footer.
+2. **No visible judgement.** Every figure is presented as though it fell out of
+   a machine. Hundreds of editorial decisions went into this codebase — which
+   metrics to show, when a benchmark is invalid, what to suppress — and not one
+   of them is visible to a reader. `isSpecialSchool()` silently drops the
+   England comparison for special schools and PRUs because that comparison is
+   meaningless; nowhere does the site *say* so.
+3. **The voice is institutional third person.** "schoolcompare brings it all
+   into one place." "Built for parents, governors, journalists." That is
+   brochure register, and it is precisely the register that machine-generated
+   content defaults to.
+
+There is a second, independent reason. The SEO programme
+(`2026-08-20-seo-programme-design.md`) defines eight workstreams and none of
+them address E-E-A-T or authorship. School performance data is YMYL territory;
+an anonymous site republishing DfE figures has no authorship signal at all. This
+work fills that hole, and the blog gives W6 (explainer content) somewhere to
+live.
+
+### The failure mode to avoid
+
+The standard fix — a stock photo and "Hi, I'm Tudor, and I'm passionate about
+education!" — reads as *more* synthetic than the current coldness. Manufactured
+warmth is a stronger machine-tell than plain institutional voice. Everything
+here has to be specific, occasionally awkward, and willing to be unflattering,
+or it makes the problem worse.
+
+## Positioning
+
+The author is **Tudor**: first name only, real photograph, no surname, no
+employer named.
+
+The credibility claim is deliberately **not** educational expertise. The About
+page states plainly: *"I'm not an education expert."* Authority comes from two
+things that are actually true:
+
+- **Experience.** A parent going through primary admissions in south-west London
+  right now. Google's E-E-A-T leads with Experience, and lived experience of the
+  thing is exactly what the DfE's own service lacks.
+- **Method.** Every number's provenance is stated, so a reader can check the
+  site rather than trust it.
+
+This is more durable than borrowed expertise: it cannot be undermined by someone
+noticing the author has no teaching qualification.
+
+**Consequence for the design.** A `Person` entity with no surname is a weak
+search signal and cannot be corroborated off-site. The credibility load
+therefore shifts onto the methodology being visibly rigorous. That is a design
+constraint, not a caveat — it is why the About page carries a substantial
+"how this is built and where it can be wrong" section rather than a short bio.
+
+### Voice rules
+
+Applied to About and every post. Recorded here so the voice does not drift.
+
+- First person singular. "I built", not "we provide".
+- Concrete over general. "when we were looking at schools in Wandsworth" beats
+  any amount of stated warmth.
+- State limits before someone else finds them. Every post that presents a
+  metric says what it does not show.
+- No mission statements, no "passionate about", no invented team.
+- No em dashes. One of the clearest tells of machine-written prose, which is
+  the exact problem this work exists to fix.
+- Short sentences. The existing code comments in this repo are already written
+  this way; the prose should match.
+
+## Scope
+
+**In:**
+
+- `/about` — a coded page (not CMS-managed).
+- `/blog` and `/blog/[slug]` — Payload-backed, with an index and post pages.
+- Payload CMS installed into the existing Next application.
+- Footer and navigation links to both.
+- `Person`, `Organization`, `BlogPosting`, `BreadcrumbList` JSON-LD.
+- RSS feed and sitemap integration.
+- One first post, so the blog does not launch empty.
+
+**Out (deliberately):**
+
+- Rewriting existing homepage/how-it-works copy into first person. Worth doing,
+  but it would double the review surface of this PR. Separate change.
+- In-product signed notes on school pages (the "distributed humanity" idea).
+  Revisit once About and the blog exist.
+- Comments, newsletter, author accounts beyond one.
+- A team page. There is no team.
+
+## Architecture
+
+### Topology
+
+Payload 3 installs **into the existing Next application** and serves `/admin`
+from the same container. One image, one deploy, no new service. This is
+Payload 3's native model and it makes on-demand revalidation trivial, because
+the CMS hooks run in the same process as the Next cache.
+
+Accepted costs: the public site's image now carries Payload, so a CMS security
+patch redeploys the whole site; and the image grows substantially.
+
+### Two collisions that must be handled
+
+**1. `/api` is already taken.** `app/api/[...path]/route.ts` is a catch-all that
+proxies `/api/*` to FastAPI at runtime. Payload's default API route is also
+`/api`. Left alone, these fight, and the failure is not clean — the catch-all
+would swallow Payload's admin API calls and forward them to FastAPI.
+
+Payload's API route is therefore remapped:
+
+```ts
+routes: { api: '/cms-api', admin: '/admin' }
+```
+
+with its route group at `app/(payload)/cms-api/[...slug]/route.ts`. The
+`/cms-api` prefix must also be added to the FastAPI proxy's excluded-paths list
+as a defensive second line.
+
+**2. `next.config.js` is CommonJS.** Payload's `withPayload()` wrapper is ESM
+only. The config must become `next.config.mjs`, converting `module.exports` to
+`export default` and wrapping the export. All existing content — the standalone
+output, `outputFileTracingIncludes`, the staging `X-Robots-Tag` header block,
+the CSP — carries over unchanged. This is mechanical but it touches the file
+that controls staging's noindex, so it needs care and an explicit test.
+
+### Database
+
+Payload uses the existing `sc_database` Postgres instance, in its **own
+`payload` schema**:
+
+```ts
+db: postgresAdapter({
+  pool: { connectionString: process.env.DATABASE_URL },
+  schemaName: 'payload',
+})
+```
+
+The frontend container is already on the `backend` Docker network, so it can
+reach `sc_database:5432` with no networking change. It needs a new
+`DATABASE_URL` environment variable.
+
+Schema isolation is not cosmetic. `public` currently holds the application
+tables and Airflow's metadata, and `scripts/migrate_csv_to_db.py --drop` exists
+to drop and reimport. Blog content living in its own schema means no data
+pipeline operation can destroy it.
+
+**Verified 2026-09-02** (this was an open question when the spec was written).
+`--drop` calls `run_full_migration()` in `backend/migration.py`, which drops
+exactly two tables by name:
+
+```python
+ks2_tables = ["school_results", "schools"]
+for tname in ks2_tables:
+    if tname in existing:
+        Base.metadata.tables[tname].drop(bind=engine)
+```
+
+There is no `Base.metadata.drop_all()` anywhere in `backend/`, and no
+`DROP SCHEMA`. The only other drop is `_apply_schema_drops()`, a single
+schema-qualified `DROP TABLE IF EXISTS marts.fact_parent_view CASCADE`.
+Nothing sets `search_path`, so the SQLAlchemy metadata resolves to `public`,
+and `inspector.get_table_names()` does not even enumerate other schemas.
+
+So the guarantee is stronger than schema isolation alone: `--drop` targets two
+named tables that Payload does not have, and would not reach `posts`, `media`
+or `users` even if they shared a schema. The `payload` schema remains the right
+choice — it protects against a *future* broadening of that script rather than
+today's behaviour — but the safety claim rests on verified code, not on
+assumption.
+
+Putting CMS tables in this instance is consistent with existing practice —
+Airflow already stores its metadata there.
+
+### Migrations
+
+Payload's Postgres adapter auto-pushes schema in development and requires
+explicit migrations in production. Use `prodMigrations`, which runs pending
+migrations during server initialisation:
+
+```ts
+db: postgresAdapter({ /* ... */, prodMigrations: migrations })
+```
+
+This is preferred over a one-shot init container (the `airflow-init` pattern)
+because the app is a single long-running process and there is no ordering
+problem to solve. Migration files are generated with `payload migrate:create`
+and committed, so schema changes travel through the same PR and staging gate as
+code.
+
+### Media
+
+Uploads go to a Docker named volume, consistent with `postgres_data`,
+`typesense_data` and `airflow_logs`.
+
+- `staticDir` must be an **absolute** path in Payload 3: `/app/media`.
+- The container runs as `nextjs` (uid 1001). The Dockerfile must
+  `mkdir -p /app/media && chown nextjs:nodejs /app/media` **before** the volume
+  is mounted, or Docker will create the mountpoint root-owned and every upload
+  will fail with EACCES.
+- `sharp` moves from `devDependencies` to `dependencies` — Payload needs it at
+  runtime to generate `imageSizes`.
+- The volume must be added to the backup routine alongside Postgres. A blog
+  post's images are not reproducible from the pipeline.
+
+### Rendering
+
+**Constraint:** CI builds the image with no database reachable. Blog pages
+therefore cannot use build-time `generateStaticParams` — that would either fail
+the build or bake in an empty post list.
+
+Instead: ISR. Post and index pages declare a `revalidate` window and render on
+first request, with Payload `afterChange` / `afterDelete` hooks calling
+`revalidatePath('/blog')` and `revalidatePath('/blog/' + slug)` for immediate
+publication. Because Payload runs in the same process, the hook calls
+`revalidatePath` from `next/cache` directly — no webhook, no shared secret.
+
+The ISR cache lives on container disk and is cleared by a redeploy. For a
+single container serving a handful of posts this is fine.
+
+### Collections
+
+- **`posts`** — `title`, `slug`, `publishedAt`, `excerpt`, `heroImage`
+  (relation to `media`), `content` (Lexical rich text), `seo` group
+  (`metaTitle`, `metaDescription`), `_status` (drafts enabled).
+- **`media`** — upload collection, `alt` required, `imageSizes` for thumbnail
+  and hero widths, public read access.
+- **`users`** — Payload's auth collection. One account. Public creation
+  disabled.
+
+Drafts are enabled so posts can be written over several sittings and previewed
+before publication.
+
+**Payload Blocks** are how posts embed live product components — a real trend
+chart or comparison table inside a post, rendered from live data rather than
+screenshotted. This is the main thing the CMS has to earn back against
+file-based MDX, and it directly serves the goal: showing judgement in context.
+Ship with one block (a callout/aside for "what this number doesn't tell you");
+add a live-chart block once a post needs it.
+
+### Security
+
+`/admin` is the first authenticated surface on this site. Public, hardened:
+
+- `PAYLOAD_SECRET` — long, random, set in the Portainer stack environment, never
+  committed. The same variable must exist in staging with a *different* value.
+- Strong unique password on the single admin account.
+- Login rate limiting via Payload's `maxLoginAttempts` / `lockTime`.
+- `X-Robots-Tag: noindex, nofollow` on `/admin/*` and `/cms-api/*`, and a
+  `robots.ts` disallow. The admin panel must never be indexed.
+- Public user creation disabled; no open registration.
+- Verify the existing CSP `frame-ancestors` directive does not break the admin
+  panel.
+
+Residual risk, accepted: a future Payload authentication CVE is live against the
+public internet. Mitigation is prompt patching, which the staging→prod pipeline
+already supports. If this becomes uncomfortable, restricting `/admin` at the
+proxy to LAN/VPN is a one-line change later.
+
+Staging note: staging runs the same image on `stx.`, so it gets its own admin
+panel and its own database. It must have its own `PAYLOAD_SECRET` and its own
+credentials — never production's.
+
+## Deployment changes
+
+- `nextjs-app/Dockerfile` — create and chown `/app/media`; ensure Payload's
+  admin bundle and `sharp` survive standalone output file tracing.
+- `docker-compose.portainer.yml` and the staging equivalent — add
+  `DATABASE_URL` and `PAYLOAD_SECRET` to the `frontend` service, add a
+  `payload_media` volume mounted at `/app/media`, and add
+  `depends_on: sc_database`.
+- Document both new environment variables in the compose header comment block,
+  which is where this stack records its configuration.
+
+## SEO
+
+- `Person` (Tudor, with photo) and `Organization` JSON-LD on `/about`.
+- `BlogPosting` + `BreadcrumbList` on post pages, with `author` referencing the
+  same `Person`.
+- Canonical URLs on `/blog` and every post.
+- Posts and `/about` added to the existing sitemap (`app/sitemap.xml/route.ts`
+  and `app/sitemaps/[...parts]`). Post URLs come from Payload at request time.
+- RSS feed at `/blog/rss.xml`.
+- Footer links to both pages, under a new "About" column.
+
+**Navigation is deliberately left alone.** `Navigation.tsx` renders a bottom tab
+bar on mobile that already carries four items (Search, Compare, Rankings,
+Admissions). A fifth tab makes each one cramped at 320px, and About and Blog are
+both lower-intent than any of the four. Both live in the footer; About
+additionally gets a byline link from every post, which is where a reader who
+cares actually asks the question. Revisit only if analytics show people hunting
+for it.
+
+## Testing
+
+Unit (Jest):
+
+- Post rendering, including a post with no hero image and one with no excerpt.
+- Slug generation and collision handling.
+- JSON-LD shape for `BlogPosting` and `Person`.
+- The `next.config.mjs` conversion preserves the staging `X-Robots-Tag` rule —
+  this guards the riskiest mechanical change in the plan.
+
+E2E (Playwright, `e2e/`, required by CLAUDE.md for user-facing change):
+
+- `/about` renders, shows the author name and photo, and is reachable from the
+  footer and nav.
+- `/blog` lists at least one post; clicking through reaches the post.
+- A post page renders title, date, body and byline.
+- `/admin` responds with `noindex` and does not leak a stack trace when
+  unauthenticated.
+
+Note the known constraint: new journeys cannot be proven in PR checks, because
+the staging E2E gate runs post-merge.
+
+## Risks
+
+| Risk | Mitigation |
+|---|---|
+| `next.config.mjs` conversion silently drops the staging noindex header, making staging a crawlable duplicate | Unit test asserting the header rule; verify on staging before promotion |
+| Payload API route collides with the FastAPI `/api` proxy | Remap to `/cms-api`; add to the proxy's exclusion list |
+| Media volume mounts root-owned; all uploads fail with EACCES | `mkdir`+`chown` in the Dockerfile before the mount; test an upload on staging |
+| Build fails or bakes empty content because CI has no DB | No build-time DB access; ISR only |
+| A pipeline `--drop` destroys blog content | Separate `payload` schema; verify `--drop` blast radius before building |
+| Media volume not backed up; images unrecoverable | Add `payload_media` to the backup routine |
+| Payload auth CVE exposed publicly | Prompt patching; proxy restriction available as a fallback |
+| Blog launches empty or goes stale | Ship with one post; cadence is explicitly "a few times a year", so no cadence is promised anywhere on the page — no dates implying a schedule |
+
+## Sequence
+
+Each step is independently reviewable and mergeable.
+
+1. **Payload foundation** — install, `next.config.mjs` conversion, `payload`
+   schema, `/cms-api` remap, `users` collection, `/admin` hardening, compose and
+   Dockerfile changes. No public-facing change yet. Verify on staging that the
+   site is unchanged and `/admin` works.
+2. **`/about`** — coded page, photo, `Person`/`Organization` JSON-LD, footer and
+   nav links, e2e journey. Independently valuable and does not depend on the
+   blog.
+3. **Blog** — `posts` and `media` collections, `/blog` index and post pages, ISR
+   plus revalidation hooks, RSS, sitemap, structured data, e2e journeys.
+4. **First post** — written in the admin panel, published through the normal
+   flow, proving the whole path end to end.
+
+Step 1 carries all the infrastructure risk and none of the visible benefit, so
+it should be verified on staging carefully before step 2 starts.
+
+## Dependencies on Tudor
+
+- **A photograph.** Blocks step 2. Nothing else in the plan is blocked by it.
+- **The first post's subject.** Blocks step 4 only. Suggested: what school
+  performance data cannot tell you — it demonstrates judgement, is genuinely
+  useful, and is the kind of thing an anonymous or machine-written site will not
+  publish.
+- ~~Confirmation that `scripts/migrate_csv_to_db.py --drop` is schema-scoped.~~
+  **Resolved 2026-09-02** — verified in `backend/migration.py`; see the
+  Database section. No action needed.

e2e/tests/journeys.spec.ts

@@ -1226,6 +1226,27 @@ const CUTOFF_CANDIDATE_URNS = [
   101099, 100553, 102574, 100769,                 // mixed
 ];
 
+/**
+ * Whether the last-distance-offered feature is switched on here.
+ *
+ * Read from the data rather than from /api/flags, which the public proxy
+ * denies on purpose — the endpoint names unreleased features. The observable
+ * effect is the field's presence: the flag is off iff no candidate school
+ * carries an `admission_distance` key at all.
+ *
+ * The distinction that matters: `admission_distance: null` means this school
+ * has no published cut-off, and the key being ABSENT means cut-offs are not
+ * being published at all.
+ */
+async function distanceFeatureIsOn(page: Page): Promise<boolean> {
+  for (const urn of CUTOFF_CANDIDATE_URNS) {
+    const res = await page.request.get(`/api/schools/${urn}`);
+    if (!res.ok()) continue;
+    if ('admission_distance' in (await res.json())) return true;
+  }
+  return false;
+}
+
 async function schoolWithCutoff(page: Page) {
   for (const urn of CUTOFF_CANDIDATE_URNS) {
     const res = await page.request.get(`/api/schools/${urn}`);
@@ -1237,6 +1258,105 @@ async function schoolWithCutoff(page: Page) {
   return null;
 }
 
+test('when the distance feature is on, a school with a cut-off is findable', async ({ page }) => {
+  /*
+   * The gate that stops the other distance journeys passing vacuously.
+   *
+   * They all skip when schoolWithCutoff() finds nothing, which is right when
+   * the feature is off — but it means a feature that is *supposed* to be on
+   * and is silently broken shows up as a green run full of skips. This test
+   * fails in that case.
+   */
+  test.skip(!(await distanceFeatureIsOn(page)),
+    'the admission_distance flag is off in this environment');
+
+  expect(await schoolWithCutoff(page),
+    'the distance feature is on, but no candidate school has a cut-off — '
+    + 'the flag is on and the data or the query behind it is broken')
+    .not.toBeNull();
+});
+
+test('with the distance feature off, the section is absent rather than empty', async ({ page }) => {
+  // Shipping dark means the page renders as it did before the feature existed,
+  // not as a feature with its content removed.
+  test.skip(await distanceFeatureIsOn(page),
+    'the admission_distance flag is on in this environment');
+
+  // A school that exists, found rather than hardcoded — a 404 page would
+  // satisfy the absent-heading assertion without proving anything.
+  //
+  // A plain loop, not Array.find: find's predicate is synchronous, so an async
+  // one returns a Promise, every Promise is truthy, and it would always hand
+  // back the first URN whether or not that school exists.
+  let urn: number | null = null;
+  for (const candidate of CUTOFF_CANDIDATE_URNS) {
+    if ((await page.request.get(`/api/schools/${candidate}`)).ok()) {
+      urn = candidate;
+      break;
+    }
+  }
+  expect(urn, 'no candidate school resolves in this environment').not.toBeNull();
+
+  await page.goto(`/school/${urn}`);
+  await expect(page.locator('h1')).toBeVisible();
+
+  await expect(page.getByRole('heading', { name: /How far away are you\?/ }))
+    .toHaveCount(0);
+});
+
+/**
+ * A secondary school carrying an EES admissions row, which is what makes its
+ * Admissions section render while the distance feature is dark.
+ */
+async function secondarySchoolWithAdmissions(page: Page) {
+  const list = await page.request.get('/api/schools?phase=secondary&page_size=40');
+  if (!list.ok()) return null;
+  const body = await list.json();
+  for (const s of (body?.schools ?? []).slice(0, 25)) {
+    const res = await page.request.get(`/api/schools/${s.urn}`);
+    if (!res.ok()) continue;
+    const detail = await res.json();
+    if (detail?.admissions == null) continue;
+    return { urn: s.urn as number };
+  }
+  return null;
+}
+
+test('with the distance feature off, a secondary page makes no claim about publication', async ({ page }) => {
+  /*
+   * Shipping dark must not put words in the council's mouth. The secondary
+   * template is the only one that words the absence, and "X has not published
+   * a cut-off distance for this school" is false wherever X does publish and
+   * we are simply withholding it.
+   *
+   * This is why the API omits the key rather than sending null: absent means
+   * "cut-offs are not published at all", null means "this school has none".
+   * Only the second is a fact about the school, and only the second is sayable.
+   */
+  test.skip(await distanceFeatureIsOn(page),
+    'the admission_distance flag is on in this environment');
+
+  const found = await secondarySchoolWithAdmissions(page);
+  test.skip(found === null, 'no secondary school in the sample has an admissions row');
+
+  await page.goto(`/school/${found!.urn}`);
+  await expect(page.locator('h1').first()).toBeVisible({ timeout: 15_000 });
+
+  // The Admissions section is still there — this is not a test that the whole
+  // section vanished, which would pass for the wrong reason.
+  await expect(page.locator('#admissions')).toHaveCount(1);
+
+  await expect(page.getByText(/has not published a cut-off distance/)).toHaveCount(0);
+  await expect(page.getByText(/Contact the admissions authority/)).toHaveCount(0);
+});
+
+test('/api/flags is not reachable from the public internet', async ({ page }) => {
+  // It names every unreleased feature and whether it is on. Next reads it
+  // server-side over the Docker network; the public proxy must deny it.
+  const res = await page.request.get('/api/flags');
+  expect(res.status()).toBe(404);
+});
+
 test('a published cut-off distance is shown with the year it belongs to', async ({ page }) => {
   const found = await schoolWithCutoff(page);
   test.skip(found === null, 'no school in the sample has a published cut-off distance yet');
@@ -1649,13 +1769,29 @@ const CANONICAL_ROUTES: Array<[string, string]> = [
   ['/admissions', 'https://www.schoolcompare.co.uk/admissions'],
 ];
 
+/**
+ * Next normalises canonical URLs against `trailingSlash: false`, so the root
+ * ships as `https://www.schoolcompare.co.uk` with no slash while every other
+ * route keeps its path. Both forms address the same document, and which one
+ * Next emits is its business, not something worth pinning a test to.
+ *
+ * The first cut hardcoded the slash and failed only on the homepage — the
+ * same gap as the doubled brand: it asserted the metadata object rather than
+ * what the page actually renders.
+ */
+function sameUrl(a: string | null, b: string): boolean {
+  const strip = (u: string) => u.replace(/\/+$/, '');
+  return strip(a ?? '') === strip(b);
+}
+
 for (const [path, expected] of CANONICAL_ROUTES) {
   test(`${path} declares exactly one canonical, on the www host`, async ({ page }) => {
     await page.goto(path);
     const hrefs = await page.locator('link[rel="canonical"]').evaluateAll(
       (els) => els.map((e) => e.getAttribute('href')));
     expect(hrefs, `${path} should declare one canonical`).toHaveLength(1);
-    expect(hrefs[0]).toBe(expected);
+    expect(sameUrl(hrefs[0], expected),
+      `${path} canonical was ${hrefs[0]}, expected ${expected}`).toBe(true);
   });
 }
 
@@ -1663,7 +1799,8 @@ test('a filtered homepage still canonicalises to the bare root', async ({ page }
   await page.goto('/?search=primary&phase=primary&sort=name&page=2');
   const href = await page.locator('link[rel="canonical"]').first()
     .getAttribute('href');
-  expect(href).toBe('https://www.schoolcompare.co.uk/');
+  expect(sameUrl(href, 'https://www.schoolcompare.co.uk/'),
+    `filtered homepage canonical was ${href}`).toBe(true);
 });
 
 test('a school page canonicalises to its own slug on the www host', async ({ page }) => {
@@ -1714,10 +1851,33 @@ test('staging answers noindex, and stays crawlable so the noindex is seen', asyn
   // The other half, and the reason this is one test rather than two: a
   // Disallow would stop Google fetching the page at all, so it would never
   // see the noindex above. The two only work together.
+  //
+  // Scoped to the `*` group. The first cut matched `Disallow: /` anywhere in
+  // the file and tripped over the AI-crawler groups Cloudflare injects —
+  // ClaudeBot, GPTBot, Amazonbot and friends all carry a blanket disallow,
+  // deliberately, and none of them is Googlebot.
   const robots = await (await page.request.get('/robots.txt')).text();
-  expect(robots).not.toMatch(/^\s*Disallow:\s*\/\s*$/mi);
+  expect(blocksEverything(robots, '*'),
+    'the * group must not disallow the whole site, or the noindex is never seen')
+    .toBe(false);
 });
 
+/** True when `agent`'s group in a robots.txt disallows the entire site. */
+function blocksEverything(robots: string, agent: string): boolean {
+  let current: string | null = null;
+  let blocked = false;
+  for (const raw of robots.split('\n')) {
+    const line = raw.split('#')[0].trim();
+    if (!line) continue;
+    const [key, ...rest] = line.split(':');
+    const value = rest.join(':').trim();
+    const k = key.trim().toLowerCase();
+    if (k === 'user-agent') current = value;
+    else if (current === agent && k === 'disallow' && value === '/') blocked = true;
+  }
+  return blocked;
+}
+
 test('a school page on staging is noindexed too, not just the homepage', async ({ page }) => {
   const list = await page.request.get('/api/schools?search=primary&per_page=1');
   const [first] = (await list.json()).schools ?? [];
@@ -1864,11 +2024,115 @@ test('a place page links its phase variants, and they resolve', async ({ page })
   await expect(page.locator('h1')).toContainText(new RegExp(`${phase} schools in`, 'i'));
 });
 
+/*
+ * The table shipped with one column of scores. A parent shortlisting from a
+ * town page needs to know whether a school takes their child's age, whether
+ * it is a faith school, and — for a primary — whether it has a nursery,
+ * before a percentage means anything.
+ *
+ * These assert the column headings rather than the values: nursery_provision
+ * and parliamentary_constituency are optional mart columns, and on an
+ * environment whose pipeline has not rebuilt them the API degrades them to
+ * absent. A value assertion would then fail for a data reason, not a code one.
+ */
+async function phasedPlace(page: Page, phase: 'primary' | 'secondary') {
+  const place = await firstPlaceOfKind(page, 'town');
+  const detail = await (await page.request.get(`/api/places/town/${place.slug}`)).json();
+  test.skip(!(detail.place.phases ?? []).includes(phase),
+    `no ${phase} page clears the threshold here`);
+  return place;
+}
+
+test('a primary place page names each school as well as scoring it', async ({ page }) => {
+  const place = await phasedPlace(page, 'primary');
+  await page.goto(`/schools/${place.slug}/primary`);
+  for (const heading of ['Ages', 'Religious character', 'Nursery', 'Constituency']) {
+    await expect(page.getByRole('columnheader', { name: heading, exact: true }))
+      .toBeVisible();
+  }
+  // age_range rides in on SCHOOL_COLUMNS and predates the optional columns,
+  // so it is the one attribute safe to assert a value for anywhere.
+  await expect(page.locator('table tbody td').filter({ hasText: /^\d+–\d+$/ }).first())
+    .toBeVisible();
+});
+
+test('a secondary place page does not ask about nurseries', async ({ page }) => {
+  const place = await phasedPlace(page, 'secondary');
+  await page.goto(`/schools/${place.slug}/secondary`);
+  await expect(page.getByRole('columnheader', { name: 'Ages', exact: true }))
+    .toBeVisible();
+  await expect(page.getByRole('columnheader', { name: 'Nursery', exact: true }))
+    .toHaveCount(0);
+});
+
+test('the measure stays beside the school name, not behind a swipe', async ({ page }) => {
+  // Six columns overflow a phone; .tableWrap turns that into a horizontal
+  // scroll. With the measure last, the number the page exists for is the one
+  // off the screen.
+  const place = await phasedPlace(page, 'primary');
+  await page.setViewportSize({ width: 390, height: 844 });
+  await page.goto(`/schools/${place.slug}/primary`);
+  const second = page.locator('table thead th').nth(1);
+  await expect(second).toContainText(/reading, writing/i);
+  await expect(second).toBeInViewport();
+});
+
 test('phase variants are submitted in the places sitemap', async ({ page }) => {
   const xml = await (await page.request.get('/sitemaps/places-1.xml')).text();
   expect(xml).toMatch(/\/schools\/[a-z0-9-]+\/primary</);
 });
 
+test('authority phase variants are submitted, and in their own namespace', async ({ page }) => {
+  // 302 of these were in the sitemap for weeks and every one 404'd: the spec
+  // called for the route, the plan built the bare authority page and dropped
+  // it, and the sitemap — written from the registry — kept submitting them.
+  const xml = await (await page.request.get('/sitemaps/places-1.xml')).text();
+  expect(xml).toMatch(/\/schools\/authority\/[a-z0-9-]+\/primary</);
+});
+
+test('every place link a place page emits resolves', async ({ page }) => {
+  /*
+   * The guard that was missing. Each family built its own links, so a URL
+   * shape belonging to one namespace was used by all four: an authority page
+   * offered "Primary schools in Barnet" pointing at /schools/barnet/primary,
+   * the *town*. For 87 of 151 authorities that 404'd; for the other 64 it
+   * quietly served a different set of schools under the same name.
+   *
+   * Only /schools links are followed. The per-school links are the same
+   * component the school-page journeys already cover, and there are hundreds
+   * of them on a page.
+   */
+  for (const kind of ['town', 'authority', 'outcode'] as const) {
+    const place = await firstPlaceOfKind(page, kind);
+    const prefix = kind === 'authority' ? '/schools/authority/'
+      : kind === 'outcode' ? '/schools/near/' : '/schools/';
+    await page.goto(`${prefix}${place.slug}`);
+
+    const hrefs = [...new Set(
+      await page.locator('a[href^="/schools"]').evaluateAll(
+        (els) => els.map((e) => e.getAttribute('href')!)))];
+    expect(hrefs.length, `${kind} page links no other place`).toBeGreaterThan(0);
+
+    for (const href of hrefs) {
+      const res = await page.request.get(href);
+      expect(res.status(), `${kind} page links ${href}`).toBe(200);
+    }
+  }
+});
+
+test('an outcode page offers no phase link, because no such page exists', async ({ page }) => {
+  // Nobody searches "primary schools in SW11", so the spec gives outcodes no
+  // phase route. The registry computed the variants anyway and the page
+  // linked them, putting two 404s on each of 1,720 outcode pages.
+  const place = await firstPlaceOfKind(page, 'outcode');
+  const detail = await (await page.request.get(
+    `/api/places/outcode/${place.slug}`)).json();
+  expect(detail.place.phases).toEqual([]);
+
+  await page.goto(`/schools/near/${place.slug}`);
+  await expect(page.getByRole('navigation', { name: 'By phase' })).toHaveCount(0);
+});
+
 test('no page title repeats the brand', async ({ page }) => {
   // The root layout appends '| schoolcompare' to a plain-string title. Any
   // route whose title already carries the brand must opt out with
@@ -1907,7 +2171,509 @@ test('a place straddling a boundary names every authority it sits in', async ({
   await page.goto(`/schools/near/${straddling!.slug}`);
 
   for (const a of detail.place.authorities) {
-    await expect(page.locator(`a[href="/schools/authority/${a.slug}"]`).first())
-      .toBeVisible();
+    if (a.slug) {
+      await expect(page.locator(`a[href="/schools/authority/${a.slug}"]`).first())
+        .toBeVisible();
+    } else {
+      // No page of its own — City of London and the Isles of Scilly are
+      // under the threshold. Named, deliberately not linked.
+      await expect(page.locator('header p')).toContainText(a.name);
+      await expect(page.getByRole('link', { name: a.name })).toHaveCount(0);
+    }
   }
 });
+
+test('a place page lists its schools alphabetically', async ({ page }) => {
+  // Someone on a place page is usually looking for a school they can name,
+  // so the order should serve scanning for it. /rankings is where the
+  // league-table ordering lives.
+  const { places } = await (await page.request.get('/api/places')).json();
+  const town = places.find((p: { kind: string; count: number }) =>
+    p.kind === 'town' && p.count >= 5);
+  expect(town).toBeTruthy();
+
+  await page.goto(`/schools/${town.slug}`);
+
+  /*
+   * Per table, not per page.
+   *
+   * An unphased place page renders one table per phase, and an all-through
+   * school legitimately appears in both — so the page's school links are not
+   * one alphabetical run and never were. This assertion used to collect them
+   * all together and only passed because no town it picked happened to hold an
+   * all-through school; when the data gave Abbots Langley one, Breakspeare
+   * School showed up in the primary table and again in the secondary, and the
+   * test failed on correct behaviour.
+   */
+  const tables = page.locator('table');
+  const tableCount = await tables.count();
+  expect(tableCount).toBeGreaterThan(0);
+
+  let checked = 0;
+  for (let i = 0; i < tableCount; i++) {
+    const names = await tables.nth(i).locator('a[href^="/school/"]').allTextContents();
+    if (names.length < 2) continue;   // a one-row table says nothing about order
+    const sorted = [...names].sort((a, b) =>
+      a.toLowerCase().localeCompare(b.toLowerCase()));
+    expect(names, `table ${i + 1} is not alphabetical`).toEqual(sorted);
+    checked++;
+  }
+  expect(checked, 'no table had enough rows to check the ordering').toBeGreaterThan(0);
+});
+
+test('the rankings page still orders by score, not name', async ({ page }) => {
+  // Alphabetical is a place-page decision, not a site-wide one.
+  const res = await page.request.get('/api/rankings?metric=rwm_expected_pct&phase=primary');
+  expect(res.ok()).toBeTruthy();
+  const scores = ((await res.json()).rankings ?? [])
+    .map((r: { rwm_expected_pct: number | null }) => r.rwm_expected_pct)
+    .filter((v: number | null) => v != null);
+  expect(scores).toEqual([...scores].sort((a: number, b: number) => b - a));
+});
+
+/*
+ * Analytics on the location layer.
+ *
+ * Umami counts a pageview for every one of these URLs already. What it cannot
+ * say is which *kind* of location page earns engagement, because all four
+ * families share the /schools/ prefix — and that is the question that decides
+ * whether to keep investing in them.
+ */
+
+/** Capture Umami events, with the real script blocked so it cannot clobber
+ *  the stub. Must be called before the first navigation. */
+async function captureEvents(page: Page) {
+  const events: Array<{ name: string; data: Record<string, unknown> }> = [];
+  await page.route('**/analytics.schoolcompare.co.uk/**', (route) => route.abort());
+  await page.exposeFunction('__capture',
+    (name: string, data: Record<string, unknown>) => { events.push({ name, data }); });
+  await page.addInitScript(() => {
+    (window as unknown as { umami: unknown }).umami = {
+      track: (name: string, data: unknown) =>
+        (window as unknown as { __capture: (n: string, d: unknown) => void })
+          .__capture(name, data),
+    };
+  });
+  return events;
+}
+
+test('a location page reports which kind of place it is', async ({ page }) => {
+  const events = await captureEvents(page);
+  const place = await firstPlaceOfKind(page, 'authority');
+
+  await page.goto(`/schools/authority/${place.slug}`);
+  await expect.poll(() => events.find((e) => e.name === 'place_viewed'),
+    { timeout: 10_000 }).toBeTruthy();
+
+  const event = events.find((e) => e.name === 'place_viewed')!;
+  expect(event.data.kind).toBe('authority');
+  expect(event.data.slug).toBe(place.slug);
+  expect(event.data.phase).toBe('all');
+});
+
+test('a school reached from a location page is attributed to it, not to direct', async ({ page }) => {
+  /*
+   * The defect this was written for. getNavigationSource had no case for
+   * /schools/, so every school view that came through the location layer was
+   * filed as 'direct' — the bucket you read as "typed the URL". The one
+   * measurement that says whether ~3,900 SEO pages work was reporting the
+   * wrong answer, confidently.
+   */
+  const events = await captureEvents(page);
+  const place = await firstPlaceOfKind(page, 'town');
+
+  await page.goto(`/schools/${place.slug}`);
+  await page.locator('a[href^="/school/"]').first().click();
+  await page.waitForURL(/\/school\//);
+
+  await expect.poll(() => events.find((e) => e.name === 'school_viewed'),
+    { timeout: 10_000 }).toBeTruthy();
+  expect(events.find((e) => e.name === 'school_viewed')!.data.from).toBe('place');
+});
+
+/*
+ * School autosuggest (spec 2026-08-26).
+ */
+async function autosuggestIsOn(page: Page): Promise<boolean> {
+  await page.goto('/');
+  return (await page.getByRole('combobox').count()) > 0;
+}
+
+test('the suggest endpoint answers from Typesense', async ({ page }) => {
+  // Not flagged — the endpoint is live even while the UI is dark, so it can
+  // be smoke-tested before the feature is switched on.
+  const res = await page.request.get('/api/suggest?q=brecknock');
+  expect(res.ok()).toBeTruthy();
+  const { suggestions } = await res.json();
+  expect(Array.isArray(suggestions)).toBeTruthy();
+  if (suggestions.length) {
+    // Local authority is what tells two "St Mary's" apart.
+    expect(suggestions[0]).toHaveProperty('school_name');
+    expect(suggestions[0]).toHaveProperty('local_authority');
+  }
+});
+
+test('a one-character query is answered, not rejected', async ({ page }) => {
+  // The keystroke path never errors on ordinary input.
+  const res = await page.request.get('/api/suggest?q=b');
+  expect(res.status()).toBe(200);
+  expect((await res.json()).suggestions).toEqual([]);
+});
+
+test('the suggest response is cacheable', async ({ page }) => {
+  const res = await page.request.get('/api/suggest?q=brecknock');
+  expect(res.headers()['cache-control'] ?? '').toContain('s-maxage');
+});
+
+test('typing a school name suggests it, and choosing it opens that school', async ({ page }) => {
+  test.skip(!(await autosuggestIsOn(page)),
+    'the school_autosuggest flag is off in this environment');
+
+  // A school certain to exist in any environment with data.
+  const { schools } = await (await page.request.get('/api/schools?page_size=1')).json();
+  test.skip(!schools?.length, 'no schools in this environment');
+  const name = schools[0].school_name as string;
+
+  await page.goto('/');
+  await page.getByRole('combobox').first().fill(name.slice(0, 12));
+  const option = page.getByRole('option').first();
+  await expect(option).toBeVisible();
+  await option.click();
+  await expect(page).toHaveURL(/\/school\/\d+/);
+});
+
+test('the whole dropdown is reachable, not clipped by the hero', async ({ page }) => {
+  /*
+   * The hero panel had overflow: hidden to clip its artwork to the rounded
+   * corners, and it clipped the dropdown too — 320px of list against 145px of
+   * panel below the input, so roughly half was cut off with nothing to say so.
+   *
+   * toBeVisible() does not catch this: it checks the box is non-empty and not
+   * visibility:hidden, and an ancestor's overflow clips neither. The invariant
+   * that does catch it is that the LAST option is the thing actually painted
+   * at its own coordinates — which fails for clipping and for occlusion alike.
+   */
+  test.skip(!(await autosuggestIsOn(page)),
+    'the school_autosuggest flag is off in this environment');
+
+  const { schools } = await (await page.request.get('/api/schools?page_size=1')).json();
+  test.skip(!schools?.length, 'no schools in this environment');
+
+  await page.goto('/');
+  await page.getByRole('combobox').first().fill(
+    (schools[0].school_name as string).slice(0, 6));
+
+  const options = page.getByRole('option');
+  await expect(options.first()).toBeVisible();
+  const count = await options.count();
+
+  const painted = await options.nth(count - 1).evaluate((el) => {
+    const r = el.getBoundingClientRect();
+    const hit = document.elementFromPoint(r.left + r.width / 2, r.top + r.height / 2);
+    return { inside: el.contains(hit) || el === hit, bottom: Math.round(r.bottom) };
+  });
+  expect(painted.inside,
+    `the last option is not painted at its own coordinates (bottom ${painted.bottom}) `
+    + '— an ancestor is clipping or covering the dropdown').toBeTruthy();
+});
+
+test('the dropdown does not survive into the results it produced', async ({ page }) => {
+  /*
+   * The bug that took the staging gate down, and it was not a test problem:
+   * after a search the results-page bar still holds the term, so the dropdown
+   * reopened on top of the results and swallowed the click on the first one.
+   * Playwright reported it as "<li role=option> intercepts pointer events"; a
+   * reader would simply have found their first result unclickable.
+   */
+  test.skip(!(await autosuggestIsOn(page)),
+    'the school_autosuggest flag is off in this environment');
+
+  await page.goto('/');
+  await page.getByRole('combobox').first().fill('school');
+  await expect(page.getByRole('option').first()).toBeVisible();
+
+  await page.getByRole('button', { name: /Search/i }).first().click();
+  await page.waitForURL(/search=school/);
+
+  await expect(page.getByRole('listbox')).toHaveCount(0);
+  // And the results underneath are actually reachable, which is the point.
+  await page.locator('a[href^="/school/"]').first().click({ timeout: 15_000 });
+  await expect(page).toHaveURL(/\/school\//);
+});
+
+test('with autosuggest off, the search box is a plain input', async ({ page }) => {
+  test.skip(await autosuggestIsOn(page),
+    'the school_autosuggest flag is on in this environment');
+
+  await page.goto('/');
+  await expect(page.getByRole('combobox')).toHaveCount(0);
+  // And the box still works: the existing search must be untouched.
+  await page.getByPlaceholder(/School name or postcode/i).first().fill('abbey');
+  await page.getByRole('button', { name: /Search/i }).first().click();
+  await expect(page).toHaveURL(/search=abbey/);
+});
+
+// ── Destination measures ───────────────────────────────────────────────────
+//
+// Two failure modes have to be told apart here, and conflating them is how
+// this suite would either hide a regression or block the promotion pipeline:
+//
+//   * the backend does not serve the `destinations` field at all — a code
+//     regression, or a deploy that did not land. FAILS.
+//   * the field is served but every school is empty — the annual EES DAG has
+//     not run on this environment yet. SKIPS, loudly.
+//
+// The second is a data-load precondition, not a defect, and it is true for
+// every commit between this merging and the DAG being triggered. Failing on it
+// would redden the staging gate for unrelated work. This is not the quiet skip
+// 4f01fbd removed from the distance journeys: that one hid a broken feature
+// behind a flag check, whereas the assertion that the code is deployed and
+// correctly shaped still runs here on every commit.
+
+async function secondaryWithDestinations(page: Page): Promise<{
+  urn: string; destinations: any;
+}> {
+  const res = await page.request.get('/api/schools?search=school&per_page=100');
+  expect(res.ok()).toBeTruthy();
+  const body = await res.json();
+  const urns: string[] = (body.schools ?? [])
+    .filter((s: { phase?: string; attainment_8_score?: number | null }) =>
+      s.phase === 'Secondary' && s.attainment_8_score != null)
+    .map((s: { urn: number }) => String(s.urn));
+  expect(urns.length).toBeGreaterThan(0);
+
+  let served = false;
+  for (const urn of urns.slice(0, 25)) {
+    const detail = await page.request.get(`/api/schools/${urn}`);
+    if (!detail.ok()) continue;
+    const data = await detail.json();
+    // The key must exist, even as null. Its absence means the backend in front
+    // of us does not know about destinations at all.
+    if ('destinations' in data) served = true;
+    if (data.destinations?.ks4) return { urn, destinations: data.destinations };
+  }
+
+  expect(served,
+    'GET /api/schools/{urn} served no `destinations` key at all — the backend '
+    + 'is missing this feature, not merely missing its data').toBeTruthy();
+
+  test.skip(true,
+    'No school has destination data yet: the annual EES DAG has not run on '
+    + 'this environment. The API shape is correct, so this is a data-load '
+    + 'precondition rather than a regression.');
+  throw new Error('unreachable');
+}
+
+test('a secondary school page says where its Year 11 leavers went', async ({ page }) => {
+  const { urn } = await secondaryWithDestinations(page);
+  await page.goto(`/school/${urn}`);
+
+  const section = page.locator('#destinations');
+  await expect(section).toBeVisible({ timeout: 15_000 });
+  await expect(section.getByRole('heading', { name: 'After Year 11' })).toBeVisible();
+  // The section must date its own cohort: destinations run about two GCSE
+  // years behind the results above them, and an undated figure reads as stale.
+  await expect(section).toContainText(/20\d{2}\/\d{2}/);
+});
+
+test('the destinations bar is absent entirely whenever a figure is withheld', async ({ page }) => {
+  const { urn, destinations } = await secondaryWithDestinations(page);
+  await page.goto(`/school/${urn}`);
+  const section = page.locator('#destinations');
+  await expect(section).toBeVisible({ timeout: 15_000 });
+
+  const allGroup = destinations.ks4.groups.all;
+  const suppressed = (allGroup?.categories ?? [])
+    .filter((c: { status: string }) => c.status === 'suppressed');
+
+  if (suppressed.length > 0) {
+    // R1: a bar drawn from the published segments leaves a gap whose width is
+    // the withheld figure, readable straight off the axis.
+    await expect(section.locator('[data-destination-segment]')).toHaveCount(0);
+    await expect(section.getByText(/withheld/i).first()).toBeVisible();
+  } else {
+    const published = (allGroup?.categories ?? [])
+      .filter((c: { status: string }) => c.status === 'published');
+    await expect(section.locator('[data-destination-segment]'))
+      .toHaveCount(published.length);
+  }
+});
+
+test('switching to disadvantaged pupils never reveals a withheld figure', async ({ page }) => {
+  const { urn, destinations } = await secondaryWithDestinations(page);
+  const disadvantaged = destinations.ks4.groups.disadvantaged;
+  test.skip(!disadvantaged, 'this school publishes no disadvantaged breakdown');
+
+  await page.goto(`/school/${urn}`);
+  const section = page.locator('#destinations');
+  await expect(section).toBeVisible({ timeout: 15_000 });
+
+  const radio = section.getByRole('radio', { name: /disadvantaged/i });
+  await expect(radio).toBeVisible();
+  await radio.click();
+
+  const suppressed = (disadvantaged.categories ?? [])
+    .filter((c: { status: string }) => c.status === 'suppressed');
+  if (suppressed.length > 0) {
+    await expect(section.locator('[data-destination-segment]')).toHaveCount(0);
+
+    // The residual must appear nowhere on the page — it is the withheld figure.
+    const cohort: number = disadvantaged.cohort;
+    const publishedTotal = (disadvantaged.categories ?? [])
+      .filter((c: { status: string }) => c.status === 'published')
+      .reduce((sum: number, c: { pupils: number }) => sum + c.pupils, 0);
+    const residual = cohort - publishedTotal;
+    const text = (await section.textContent()) ?? '';
+    expect(text).not.toMatch(new RegExp(`\\b${residual}\\b`));
+  }
+});
+
+test('a school with no sixth form has no post-16 destinations section', async ({ page }) => {
+  const res = await page.request.get('/api/schools?search=school&per_page=100');
+  const body = await res.json();
+  const noSixthForm = (body.schools ?? [])
+    .filter((s: { phase?: string; has_sixth_form?: boolean }) =>
+      s.phase === 'Secondary' && s.has_sixth_form === false)
+    .map((s: { urn: number }) => String(s.urn));
+  test.skip(noSixthForm.length === 0, 'no sixth-form-less secondary in this dataset');
+
+  await page.goto(`/school/${noSixthForm[0]}`);
+  await expect(page.locator('h1').first()).toBeVisible({ timeout: 15_000 });
+  // Absence is the correct statement, so there must be no placeholder either.
+  await expect(page.locator('#post16-destinations')).toHaveCount(0);
+  await expect(page.getByText(/destination data coming soon/i)).toHaveCount(0);
+});
+
+test('the destinations section never claims a pupil stayed at this school', async ({ page }) => {
+  const { urn } = await secondaryWithDestinations(page);
+  await page.goto(`/school/${urn}`);
+  const section = page.locator('#destinations');
+  await expect(section).toBeVisible({ timeout: 15_000 });
+  // The published file records the TYPE of place a leaver went to, never which
+  // one, so the page can never say a pupil stayed on here.
+  const text = (await section.textContent()) ?? '';
+  expect(text).not.toMatch(/stayed on (here|at this school)/i);
+});
+
+/**
+ * The About page and the blog exist to give the site a named human author.
+ * These journeys assert the load-bearing parts of that — a name, a face, the
+ * honesty claim, and a resolvable Person entity — rather than exact copy,
+ * which will be edited.
+ *
+ * Both are behind flags (about_page, blog), so each has a lit journey and a
+ * dark one. Flag state is read from the observable effect rather than from
+ * /api/flags, which the public proxy denies on purpose — the same approach
+ * distanceFeatureIsOn() takes above.
+ */
+async function aboutPageIsOn(page: Page): Promise<boolean> {
+  return (await page.request.get('/about')).ok();
+}
+
+async function blogIsOn(page: Page): Promise<boolean> {
+  return (await page.request.get('/blog')).ok();
+}
+
+test('with the about page off, it is absent rather than empty', async ({ page }) => {
+  test.skip(await aboutPageIsOn(page), 'the about_page flag is on in this environment');
+
+  // Dark means the URL does not exist, not that it renders empty: a 404 is
+  // what stops a crawler keeping the page in its index.
+  expect((await page.request.get('/about')).status()).toBe(404);
+
+  // A footer link into a 404 is the failure this flag has to avoid.
+  await page.goto('/');
+  await expect(page.locator('footer a[href="/about"]')).toHaveCount(0);
+
+  // And a sitemap must never advertise a URL that 404s.
+  const sitemap = await page.request.get('/content-sitemap.xml');
+  expect(await sitemap.text()).not.toContain('/about');
+});
+
+test('with the blog off, it is absent rather than empty', async ({ page }) => {
+  test.skip(await blogIsOn(page), 'the blog flag is on in this environment');
+
+  expect((await page.request.get('/blog')).status()).toBe(404);
+  expect((await page.request.get('/blog/rss.xml')).status()).toBe(404);
+
+  await page.goto('/');
+  await expect(page.locator('footer a[href="/blog"]')).toHaveCount(0);
+
+  const sitemap = await page.request.get('/content-sitemap.xml');
+  expect(await sitemap.text()).not.toContain('/blog');
+
+  // The admin panel is deliberately NOT flagged: posts have to be writable
+  // before the blog is readable, or there is nothing to turn on.
+  expect((await page.request.get('/admin')).status()).not.toBe(404);
+});
+
+test('the about page names a human author and is reachable from the footer', async ({ page }) => {
+  test.skip(!(await aboutPageIsOn(page)), 'the about_page flag is off in this environment');
+
+  await page.goto('/');
+  const aboutLink = page.locator('footer a[href="/about"]');
+  await expect(aboutLink).toBeVisible();
+  await aboutLink.click();
+  await page.waitForURL(/\/about$/);
+
+  await expect(page.getByRole('heading', { level: 1 })).toContainText('Tudor');
+  await expect(page.locator('img[alt*="Tudor"]')).toBeVisible();
+
+  // The credibility claim is lived experience plus stated provenance, not
+  // expertise. If this sentence ever disappears the positioning has drifted.
+  await expect(page.getByText(/not an education expert/i)).toBeVisible();
+
+  const jsonLd = await page
+    .locator('script[type="application/ld+json"]')
+    .first()
+    .textContent();
+  expect(jsonLd).toContain('"Person"');
+  // First name only — a surname here would be the one place it leaks.
+  expect(jsonLd).not.toMatch(/familyName/);
+});
+
+test('the blog lists posts and each one renders with a byline', async ({ page }) => {
+  test.skip(!(await blogIsOn(page)), 'the blog flag is off in this environment');
+
+  await page.goto('/blog');
+  await expect(page.getByRole('heading', { level: 1 })).toBeVisible();
+
+  const postLinks = page.locator('a[href^="/blog/"]');
+  // Data invariant: staging must carry at least one published post. If this
+  // fails, the environment has no content rather than the code being broken.
+  expect(await postLinks.count()).toBeGreaterThan(0);
+
+  await postLinks.first().click();
+  await page.waitForURL(/\/blog\/.+/);
+  await expect(page.getByRole('heading', { level: 1 })).toBeVisible();
+  await expect(page.getByText(/^By Tudor/)).toBeVisible();
+
+  const jsonLd = await page
+    .locator('script[type="application/ld+json"]')
+    .first()
+    .textContent();
+  expect(jsonLd).toContain('"BlogPosting"');
+});
+
+test('the admin panel is not indexable', async ({ page }) => {
+  const response = await page.request.get('/admin');
+  expect(response.headers()['x-robots-tag']).toContain('noindex');
+});
+
+test('the content sitemap lists the about page and is advertised in robots', async ({ page }) => {
+  const sitemap = await page.request.get('/content-sitemap.xml');
+  // Served whatever the flags say: robots.txt names it unconditionally, and
+  // with both dark it is a valid empty urlset rather than a 404.
+  expect(sitemap.ok()).toBeTruthy();
+
+  if (await aboutPageIsOn(page)) {
+    expect(await sitemap.text()).toContain('/about');
+  }
+
+  // The school corpus sitemap is proxied from FastAPI; this one is Next's.
+  // robots.txt must advertise both or the blog never gets discovered.
+  const robots = await page.request.get('/robots.txt');
+  const body = await robots.text();
+  expect(body).toContain('/sitemap.xml');
+  expect(body).toContain('/content-sitemap.xml');
+});

nextjs-app/.gitignore

@@ -39,3 +39,4 @@ yarn-error.log*
 # typescript
 *.tsbuildinfo
 next-env.d.ts
+

nextjs-app/Dockerfile

@@ -53,6 +53,13 @@ COPY --from=builder /app/.next/static ./.next/static
 # a miss here is a silent 500 on /opengraph-image, not a build failure.
 COPY --from=builder /app/assets ./assets
 
+# Payload writes uploads here, and the compose file mounts a named volume over
+# it. The directory must exist and be owned by the runtime user BEFORE the
+# mount: Docker seeds a fresh named volume from the image path, so a missing or
+# root-owned directory here makes every upload fail with EACCES at runtime,
+# long after the build passed. The chown below covers it.
+RUN mkdir -p /app/media
+
 # Set correct permissions
 RUN chown -R nextjs:nodejs /app
 

nextjs-app/__tests__/api/proxyDenylist.test.ts

@@ -0,0 +1,31 @@
+/**
+ * The /api/* proxy is public. Anything it forwards is on the internet.
+ *
+ * @jest-environment node
+ */
+// The docblock above is load-bearing. jest.config.js sets jsdom globally, and
+// NextRequest/NextResponse need the Web Fetch API globals that only the node
+// environment provides — under jsdom this suite fails on import, not on an
+// assertion.
+import { NextRequest } from 'next/server';
+import { GET } from '@/app/(frontend)/api/[...path]/route';
+
+function request(path: string) {
+  return new NextRequest(`http://localhost:3000/api/${path}`);
+}
+
+describe('public API proxy', () => {
+  it('refuses to forward internal-only paths', async () => {
+    // /api/flags names every unreleased feature and its state. Forwarding it
+    // publishes the thing shipping dark exists to keep quiet.
+    const res = await GET(request('flags'), { params: Promise.resolve({ path: ['flags'] }) });
+    expect(res.status).toBe(404);
+  });
+
+  it('does not deny a path that merely starts with the same letters', async () => {
+    // A prefix match would take /api/flagship down with /api/flags.
+    const res = await GET(
+      request('flagship'), { params: Promise.resolve({ path: ['flagship'] }) });
+    expect(res.status).not.toBe(404);
+  });
+});

nextjs-app/__tests__/app/aboutMetadata.test.ts

@@ -0,0 +1,34 @@
+import { metadata } from '@/app/(frontend)/about/page';
+import { personJsonLd, organizationJsonLd } from '@/lib/jsonld';
+
+describe('/about metadata', () => {
+  it('canonicalises to the bare path', () => {
+    expect(metadata.alternates?.canonical)
+      .toBe('https://www.schoolcompare.co.uk/about');
+  });
+});
+
+describe('author structured data', () => {
+  it('describes a Person with a first name and a photo', () => {
+    const person = personJsonLd();
+    expect(person['@type']).toBe('Person');
+    expect(person.name).toBe('Tudor');
+    expect(person.image).toBe('https://www.schoolcompare.co.uk/brand/tudor.jpg');
+    expect(person.url).toBe('https://www.schoolcompare.co.uk/about');
+  });
+
+  it('never publishes a surname or an employer', () => {
+    // Author identity constraint: first name only. A surname here would be
+    // the one place it leaks, since JSON-LD is machine-read and archived.
+    const serialised = JSON.stringify(personJsonLd());
+    expect(serialised).not.toMatch(/familyName|Sitaru/i);
+    expect(serialised).not.toMatch(/worksFor|affiliation/i);
+  });
+
+  it('describes the site as an Organization the Person authors for', () => {
+    const org = organizationJsonLd();
+    expect(org['@type']).toBe('Organization');
+    expect(org.name).toBe('schoolcompare');
+    expect(org.url).toBe('https://www.schoolcompare.co.uk');
+  });
+});

nextjs-app/__tests__/app/blogMetadata.test.ts

@@ -0,0 +1,66 @@
+/**
+ * The blog index imports getCachedPayload, which pulls in Payload — ESM-only,
+ * and next/jest will not transform node_modules. Mocking that one module keeps
+ * the page's metadata testable without loading the CMS; the mock is never
+ * called, because `metadata` is a static export evaluated at import time.
+ */
+jest.mock('@/lib/payload', () => ({ getCachedPayload: jest.fn() }));
+
+import { metadata } from '@/app/(frontend)/blog/page';
+import { blogPostingJsonLd, breadcrumbJsonLd } from '@/lib/jsonld';
+
+const post = {
+  title: 'What the data cannot tell you',
+  slug: 'what-the-data-cannot-tell-you',
+  excerpt: 'Results describe one year group on a handful of days.',
+  publishedAt: '2026-09-15T00:00:00.000Z',
+};
+
+describe('/blog metadata', () => {
+  it('canonicalises to the bare path', () => {
+    expect(metadata.alternates?.canonical)
+      .toBe('https://www.schoolcompare.co.uk/blog');
+  });
+});
+
+describe('BlogPosting structured data', () => {
+  it('names the same Person entity the about page declares', () => {
+    // By @id, not by repeating the person: search engines must resolve every
+    // post and the about page to one author entity, or the site has several.
+    const ld = blogPostingJsonLd(post, { namedAuthor: true });
+    expect(ld['@type']).toBe('BlogPosting');
+    expect(ld.author['@id']).toBe('https://www.schoolcompare.co.uk/about#tudor');
+    expect(ld.publisher['@id']).toBe('https://www.schoolcompare.co.uk#organization');
+  });
+
+  it('attributes to the organization when the about page is dark', () => {
+    /*
+     * The two flags are independent, so blog-on-about-off is a reachable
+     * state. The Person entity lives at /about#tudor and that URL 404s while
+     * the flag is dark, so claiming it would declare an author that resolves
+     * to nothing — worse for the blog's credibility than having no named
+     * author at all. Attribute to the publisher instead.
+     */
+    const ld = blogPostingJsonLd(post, { namedAuthor: false });
+    expect(ld.author['@id']).toBe('https://www.schoolcompare.co.uk#organization');
+    expect(JSON.stringify(ld)).not.toContain('/about');
+  });
+
+  it('carries a self-referencing canonical url and the publish date', () => {
+    const ld = blogPostingJsonLd(post, { namedAuthor: true });
+    expect(ld.url).toBe(
+      'https://www.schoolcompare.co.uk/blog/what-the-data-cannot-tell-you',
+    );
+    expect(ld.datePublished).toBe('2026-09-15T00:00:00.000Z');
+  });
+});
+
+describe('breadcrumbs', () => {
+  it('places the post under the blog index', () => {
+    const ld = breadcrumbJsonLd(post);
+    expect(ld.itemListElement[0].item).toBe('https://www.schoolcompare.co.uk/blog');
+    expect(ld.itemListElement[1].item).toBe(
+      'https://www.schoolcompare.co.uk/blog/what-the-data-cannot-tell-you',
+    );
+  });
+});

nextjs-app/__tests__/app/metadata.test.ts

@@ -1,7 +1,7 @@
-import { metadata as homeMetadata } from '@/app/page';
-import { metadata as rankingsMetadata } from '@/app/rankings/page';
-import { metadata as admissionsMetadata } from '@/app/admissions/page';
-import { generateMetadata as compareMetadata } from '@/app/compare/page';
+import { metadata as homeMetadata } from '@/app/(frontend)/page';
+import { metadata as rankingsMetadata } from '@/app/(frontend)/rankings/page';
+import { metadata as admissionsMetadata } from '@/app/(frontend)/admissions/page';
+import { generateMetadata as compareMetadata } from '@/app/(frontend)/compare/page';
 
 describe('canonical URLs', () => {
   it('the homepage canonicalises to the bare root', () => {

nextjs-app/__tests__/app/nextConfig.test.ts

@@ -0,0 +1,66 @@
+/**
+ * next.config.mjs carries the staging noindex rule. Breaking it turns
+ * stx.schoolcompare.co.uk into a fully crawlable duplicate of production,
+ * and nothing else in the suite would notice.
+ *
+ * The non-null assertions are deliberate: every key asserted here is optional
+ * on NextConfig, and a missing one is precisely the regression under test, so
+ * the assertion below should fail the test rather than the compile.
+ */
+import nextConfig from '@/next.config.mjs';
+
+async function headerRules() {
+  return nextConfig.headers!();
+}
+
+describe('next.config.mjs', () => {
+  it('keeps the staging host out of the index', async () => {
+    const headers = await headerRules();
+    const stagingRule = headers.find((rule) =>
+      rule.has?.some(
+        (cond) => cond.type === 'host' && cond.value === 'stx.schoolcompare.co.uk',
+      ),
+    );
+    expect(stagingRule).toBeDefined();
+    expect(stagingRule!.headers).toContainEqual({
+      key: 'X-Robots-Tag',
+      value: 'noindex, nofollow',
+    });
+  });
+
+  it('still emits standalone output for the Docker runner', () => {
+    expect(nextConfig.output).toBe('standalone');
+  });
+
+  it('still traces the share-card fonts into the standalone bundle', () => {
+    expect(nextConfig.outputFileTracingIncludes!['/opengraph-image']).toEqual([
+      './assets/**',
+    ]);
+  });
+
+  it('still allows the analytics subdomain to frame the site', async () => {
+    const headers = await headerRules();
+    const csp = headers
+      .flatMap((rule) => rule.headers)
+      .find((header) => header.key === 'Content-Security-Policy');
+    expect(csp).toBeDefined();
+    expect(csp!.value).toContain('https://analytics.schoolcompare.co.uk');
+  });
+});
+
+describe('admin surface', () => {
+  it('serves noindex on the admin panel and the CMS API', async () => {
+    // robots.txt disallows these too, but a Disallow only blocks crawling — a
+    // URL found from an external link can still be indexed without ever being
+    // fetched. This header is what actually keeps them out.
+    const headers = await headerRules();
+    for (const source of ['/admin/:path*', '/cms-api/:path*']) {
+      const rule = headers.find((entry) => entry.source === source);
+      expect(rule).toBeDefined();
+      expect(rule!.headers).toContainEqual({
+        key: 'X-Robots-Tag',
+        value: 'noindex, nofollow',
+      });
+    }
+  });
+});

nextjs-app/__tests__/app/placeMetadata.test.ts

@@ -1,4 +1,4 @@
-import { generateMetadata as placeMeta } from '@/app/schools/[place]/page';
+import { generateMetadata as placeMeta } from '@/app/(frontend)/schools/[place]/page';
 
 jest.mock('@/lib/places', () => ({
   ...jest.requireActual('@/lib/places'),

nextjs-app/__tests__/app/robots.test.ts

@@ -0,0 +1,20 @@
+import robots from '@/app/robots';
+
+describe('robots.txt', () => {
+  it('disallows the admin panel and the CMS API', () => {
+    const rules = robots().rules;
+    const rule = Array.isArray(rules) ? rules[0] : rules;
+    expect(rule.disallow).toEqual(
+      expect.arrayContaining(['/api/', '/_next/', '/admin/', '/cms-api/']),
+    );
+  });
+});
+
+describe('sitemap discovery', () => {
+  it('lists both the proxied school sitemap and the Next-owned content sitemap', () => {
+    expect(robots().sitemap).toEqual([
+      'https://www.schoolcompare.co.uk/sitemap.xml',
+      'https://www.schoolcompare.co.uk/content-sitemap.xml',
+    ]);
+  });
+});

nextjs-app/__tests__/components/DestinationsSection.test.tsx

@@ -0,0 +1,149 @@
+import { render, screen } from '@testing-library/react';
+import { DestinationsSection } from '@/components/school/DestinationsSection';
+import type { DestinationPhase } from '@/lib/types';
+import type { DestinationCategory, DestinationStatus } from '@/lib/destinations';
+
+const cell = (
+  category: DestinationCategory,
+  pupils: number | null,
+  status: DestinationStatus = 'published',
+) => ({
+  category, pupils,
+  percentage: pupils === null ? null : (pupils / 180) * 100,
+  status,
+});
+
+const ALL_PUBLISHED = [
+  cell('school_sixth_form', 75), cell('sixth_form_college', 21),
+  cell('further_education', 55), cell('other_education', 6),
+  cell('apprenticeship', 8), cell('employment', 6),
+  cell('not_sustained', 5), cell('not_captured', 4),
+];
+
+const fullPhase: DestinationPhase = {
+  cohort_year: '2022/23',
+  groups: { all: { cohort: 180, categories: ALL_PUBLISHED } },
+};
+
+const suppressedPhase: DestinationPhase = {
+  cohort_year: '2022/23',
+  groups: {
+    all: {
+      cohort: 180,
+      categories: [
+        cell('school_sixth_form', 75), cell('sixth_form_college', null, 'suppressed'),
+        cell('further_education', 55), cell('other_education', 6),
+        cell('apprenticeship', 8), cell('employment', 6),
+        cell('not_sustained', 5), cell('not_captured', 4),
+      ],
+    },
+  },
+};
+
+describe('DestinationsSection', () => {
+  it('dates its own cohort so it is not read as stale next to the GCSE section', () => {
+    render(<DestinationsSection destinations={fullPhase} />);
+    expect(screen.getByText(/2022\/23/)).toBeInTheDocument();
+  });
+
+  it('renders one bar segment per published category', () => {
+    const { container } = render(<DestinationsSection destinations={fullPhase} />);
+    expect(container.querySelectorAll('[data-destination-segment]')).toHaveLength(8);
+  });
+
+  it('renders NO bar at all when a category is withheld', () => {
+    const { container } = render(<DestinationsSection destinations={suppressedPhase} />);
+    // R1: a bar with a gap in it publishes the withheld figure by its width.
+    expect(container.querySelectorAll('[data-destination-segment]')).toHaveLength(0);
+    expect(screen.getAllByText(/withheld/i).length).toBeGreaterThan(0);
+  });
+
+  it('never states the remainder for a partially suppressed group', () => {
+    const { container } = render(<DestinationsSection destinations={suppressedPhase} />);
+    // 180 cohort - 159 published = 21, the withheld figure. It must appear nowhere.
+    expect(container.textContent).not.toMatch(/\b21\b/);
+  });
+
+  it('shows a card value for a group whose components are all published', () => {
+    render(<DestinationsSection destinations={fullPhase} />);
+    // academic route = 75 + 21 = 96 of 180 = 53%
+    expect(screen.getByText('53%')).toBeInTheDocument();
+  });
+
+  it('refuses a card value when one of its components is withheld', () => {
+    render(<DestinationsSection destinations={suppressedPhase} />);
+    // academic route needs sixth_form_college, which is suppressed.
+    expect(screen.getByText(/not published/i)).toBeInTheDocument();
+    expect(screen.queryByText('53%')).not.toBeInTheDocument();
+  });
+
+  it('never claims a pupil stayed at this school', () => {
+    const { container } = render(<DestinationsSection destinations={fullPhase} />);
+    // The published file reports destination TYPE, never destination institution.
+    expect(container.textContent).not.toMatch(/stayed on (here|at this school)/i);
+  });
+
+  it('renders nothing when no group carries categories', () => {
+    const empty: DestinationPhase = { cohort_year: '2022/23', groups: {} };
+    const { container } = render(<DestinationsSection destinations={empty} />);
+    expect(container.firstChild).toBeNull();
+  });
+});
+
+describe('the detail table keeps the three statuses apart', () => {
+  // 'suppressed' and 'not_applicable' are different claims, and the mart, the
+  // SQLAlchemy model and the serialiser all preserve the difference. The table
+  // used to key its Share column off `percentage === null`, which is true for
+  // both, so a category that simply does not apply was labelled "withheld" —
+  // while the Pupils column beside it rendered blank.
+  const mixedPhase: DestinationPhase = {
+    cohort_year: '2022/23',
+    groups: {
+      all: {
+        cohort: 180,
+        categories: [
+          cell('school_sixth_form', 75),
+          cell('sixth_form_college', null, 'suppressed'),
+          cell('further_education', null, 'suppressed'),
+          cell('apprenticeship', null, 'not_applicable'),
+        ],
+      },
+    },
+  };
+
+  const rowFor = (container: HTMLElement, category: string) =>
+    Array.from(container.querySelectorAll('tbody tr'))
+      .find(tr => tr.textContent?.includes(category));
+
+  it('never labels a not-applicable category as withheld', () => {
+    const { container } = render(<DestinationsSection destinations={mixedPhase} />);
+    const row = rowFor(container, 'Apprenticeship');
+    expect(row).toBeTruthy();
+    expect(row!.textContent).not.toMatch(/withheld/i);
+  });
+
+  it('labels a genuinely suppressed category as withheld in both columns', () => {
+    const { container } = render(<DestinationsSection destinations={mixedPhase} />);
+    const row = rowFor(container, 'Sixth-form college');
+    expect(row).toBeTruthy();
+    expect(row!.querySelectorAll('td')).toHaveLength(2);
+    Array.from(row!.querySelectorAll('td')).forEach(td =>
+      expect(td.textContent).toMatch(/withheld/i));
+  });
+
+  it('the two columns of a row never disagree about what the row is', () => {
+    const { container } = render(<DestinationsSection destinations={mixedPhase} />);
+    Array.from(container.querySelectorAll('tbody tr')).forEach(tr => {
+      const cells = Array.from(tr.querySelectorAll('td'))
+        .map(td => /withheld/i.test(td.textContent ?? ''));
+      expect(new Set(cells).size).toBe(1);
+    });
+  });
+
+  it('shows a published category its real figures', () => {
+    const { container } = render(<DestinationsSection destinations={mixedPhase} />);
+    const row = rowFor(container, 'State-funded school sixth form');
+    expect(row!.textContent).toMatch(/75/);
+    expect(row!.textContent).toMatch(/42%/);
+  });
+});

nextjs-app/__tests__/components/FilterBarSuggest.test.tsx

@@ -0,0 +1,110 @@
+import { render, screen, fireEvent, waitFor } from '@testing-library/react';
+import userEvent from '@testing-library/user-event';
+import { FilterBar } from '@/components/FilterBar';
+
+const push = jest.fn();
+let searchParams = new URLSearchParams();
+jest.mock('next/navigation', () => ({
+  useRouter: () => ({ push, replace: jest.fn(), prefetch: jest.fn() }),
+  usePathname: () => '/',
+  useSearchParams: () => searchParams,
+}));
+
+const FILTERS = {
+  local_authorities: [], school_types: [], years: [], phases: [],
+  genders: [], admissions_policies: [],
+};
+
+const realFetch = global.fetch;
+beforeEach(() => {
+  global.fetch = jest.fn(async () => ({
+    ok: true,
+    json: async () => ({ suggestions: [{
+      urn: 100010, school_name: 'Brecknock Primary School',
+      local_authority: 'Camden', postcode: 'NW1 1AA',
+      phase: 'Primary', school_type: 'Community school' }] }),
+  })) as unknown as typeof fetch;
+  push.mockClear();
+  searchParams = new URLSearchParams();
+});
+afterEach(() => { global.fetch = realFetch; });
+
+describe('FilterBar autosuggest', () => {
+  it('is a combobox only when the flag is on', () => {
+    const { rerender } = render(<FilterBar filters={FILTERS} autosuggest={false} />);
+    expect(screen.queryByRole('combobox')).not.toBeInTheDocument();
+    rerender(<FilterBar filters={FILTERS} autosuggest />);
+    expect(screen.getByRole('combobox')).toBeInTheDocument();
+  });
+
+  it('makes no request while the flag is off', async () => {
+    // Off means off: no listener, no fetch, no markup.
+    render(<FilterBar filters={FILTERS} autosuggest={false} />);
+    await userEvent.type(screen.getByPlaceholderText(/School name or postcode/i),
+      'brecknock');
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+
+  it('shows suggestions and navigates when one is chosen', async () => {
+    render(<FilterBar filters={FILTERS} autosuggest />);
+    await userEvent.type(screen.getByRole('combobox'), 'brecknock');
+    const option = await screen.findByRole('option', { name: /Brecknock/ });
+    await userEvent.click(option);
+    expect(push).toHaveBeenCalledWith(
+      expect.stringContaining('/school/100010'));
+  });
+
+  it('suppresses suggestions once the value is a postcode', async () => {
+    // The box takes a name OR a postcode; suggestions must get out of the way.
+    //
+    // fireEvent.change, not userEvent.type: typing sets "N", "NW", "NW1"... and
+    // "NW1" is not a postcode, so a request for it is correct behaviour. Only
+    // the settled value is the assertion, so set it in one go.
+    render(<FilterBar filters={FILTERS} autosuggest />);
+    fireEvent.change(screen.getByRole('combobox'), { target: { value: 'NW1 1AA' } });
+    await new Promise((r) => setTimeout(r, 300));   // past the 200ms debounce
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+
+  it('Enter with no active option still submits the free-text search', async () => {
+    // The existing behaviour is preserved, not replaced.
+    render(<FilterBar filters={FILTERS} autosuggest />);
+    const input = screen.getByRole('combobox');
+    await userEvent.type(input, 'brecknock{Enter}');
+    // updateURL pushes inside startTransition, so the call is not synchronous.
+    await waitFor(() => expect(push).toHaveBeenCalledWith(
+      expect.stringContaining('search=brecknock')));
+  });
+});
+
+describe('FilterBar autosuggest does not reopen over results', () => {
+  it('stays shut when the input arrives pre-filled from the URL', async () => {
+    /*
+     * The results-page bar renders with the search term already in the input.
+     * Opening on that would drop the dropdown on top of the results the search
+     * just produced — which is exactly what happened: the first result became
+     * unclickable, because the list sat over it and swallowed the pointer.
+     *
+     * Suggestions answer typing, not the presence of a value.
+     */
+    searchParams = new URLSearchParams('search=brecknock');
+    render(<FilterBar filters={FILTERS} autosuggest />);
+
+    expect(screen.getByRole('combobox')).toHaveValue('brecknock');
+    await new Promise((r) => setTimeout(r, 300));   // past the 200ms debounce
+    expect(global.fetch).not.toHaveBeenCalled();
+    expect(screen.queryByRole('listbox')).not.toBeInTheDocument();
+  });
+
+  it('closes the dropdown when the search is submitted', async () => {
+    render(<FilterBar filters={FILTERS} autosuggest />);
+    const input = screen.getByRole('combobox');
+
+    await userEvent.type(input, 'brecknock');
+    expect(await screen.findByRole('listbox')).toBeInTheDocument();
+
+    await userEvent.type(input, '{Enter}');
+    await waitFor(() =>
+      expect(screen.queryByRole('listbox')).not.toBeInTheDocument());
+  });
+});

nextjs-app/__tests__/components/Footer.test.tsx

@@ -0,0 +1,47 @@
+/**
+ * The footer is the only navigational route to /about and /blog, so it is
+ * where a dark flag would otherwise leave a link into a 404.
+ *
+ * Both props default to false. A caller that forgets to pass them hides the
+ * links, which is the direction that cannot break a page — the same reasoning
+ * as backend/flags.py's "every flag defaults to False".
+ */
+import { render, screen } from '@testing-library/react';
+import { Footer } from '@/components/Footer';
+
+describe('footer feature links', () => {
+  it('links to both when both flags are on', () => {
+    render(<Footer aboutEnabled blogEnabled />);
+    expect(screen.getByRole('link', { name: /who's behind this/i }))
+      .toHaveAttribute('href', '/about');
+    expect(screen.getByRole('link', { name: /^blog$/i }))
+      .toHaveAttribute('href', '/blog');
+  });
+
+  it('omits the about link when that flag is dark', () => {
+    render(<Footer blogEnabled />);
+    expect(screen.queryByRole('link', { name: /who's behind this/i })).toBeNull();
+    expect(screen.getByRole('link', { name: /^blog$/i })).toBeInTheDocument();
+  });
+
+  it('omits the blog link when that flag is dark', () => {
+    render(<Footer aboutEnabled />);
+    expect(screen.queryByRole('link', { name: /^blog$/i })).toBeNull();
+    expect(screen.getByRole('link', { name: /who's behind this/i }))
+      .toBeInTheDocument();
+  });
+
+  it('drops the whole section when both are dark, not an empty heading', () => {
+    // Shipping dark means the footer renders as it did before the feature
+    // existed, not as a section with its contents removed.
+    render(<Footer />);
+    expect(screen.queryByRole('heading', { name: /^about$/i })).toBeNull();
+    expect(screen.queryByRole('link', { name: /who's behind this/i })).toBeNull();
+    expect(screen.queryByRole('link', { name: /^blog$/i })).toBeNull();
+  });
+
+  it('defaults to dark when a caller passes nothing', () => {
+    render(<Footer />);
+    expect(screen.queryByRole('link', { name: /who's behind this/i })).toBeNull();
+  });
+});

nextjs-app/__tests__/components/PlaceView.test.tsx

@@ -243,3 +243,237 @@ describe('PlaceView authorities', () => {
     expect(screen.getByRole('link', { name: 'Merton' })).toBeInTheDocument();
   });
 });
+
+describe('PlaceView list ordering', () => {
+  const detail3: PlaceDetail = {
+    place: { kind: 'town', slug: 'brentwood', name: 'Brentwood', count: 2,
+             parent_authority: 'Essex', phases: ['primary'] },
+    schools: [
+      { urn: 1, school_name: 'Alpha Primary', phase: 'Primary',
+        rwm_expected_pct: 40, attainment_8_score: null } as never,
+      { urn: 2, school_name: 'Beta Primary', phase: 'Primary',
+        rwm_expected_pct: 90, attainment_8_score: null } as never,
+    ],
+    averages: { rwm_expected_pct: 65, attainment_8_score: null },
+  };
+
+  it('renders schools in the order the API sent them, not by score', () => {
+    // The API sorts alphabetically now; the component must not re-sort.
+    render(<PlaceView detail={detail3} englandAverage={61} neighbours={[]} />);
+    const links = screen.getAllByRole('link', { name: /Primary$/ });
+    expect(links.map((l) => l.textContent))
+      .toEqual(['Alpha Primary', 'Beta Primary']);
+  });
+
+  it('declares the list as ascending rather than implying a ranking', () => {
+    // An ItemList carrying `position` reads as a ranking unless it says
+    // otherwise, and the table is A-Z.
+    const { container } = render(<PlaceView detail={detail3} englandAverage={61}
+      neighbours={[]} />);
+    const ld = JSON.parse(
+      container.querySelector('script[type="application/ld+json"]')!.textContent!);
+    const list = ld['@graph'].find((n: { '@type': string }) => n['@type'] === 'ItemList');
+    expect(list.itemListOrder).toBe('https://schema.org/ItemListOrderAscending');
+  });
+});
+
+describe('PlaceView phase links', () => {
+  const authority: PlaceDetail = {
+    place: { kind: 'authority', slug: 'barnet', name: 'Barnet', count: 156,
+             parent_authority: null, phases: ['primary', 'secondary'] },
+    schools: [
+      { urn: 1, school_name: 'Alpha Primary', phase: 'Primary',
+        rwm_expected_pct: 82, attainment_8_score: null } as never,
+    ],
+    averages: { rwm_expected_pct: 63, attainment_8_score: null },
+  };
+
+  it('keeps an authority phase link in the authority namespace', () => {
+    // The link was built as `/schools/${slug}/${phase}` for every kind, so an
+    // authority page pointed into the town namespace. For 87 of 151
+    // authorities that 404'd; for the other 64 it silently landed on the town
+    // page of the same name — a different set of schools, and exactly the
+    // duplicate the two namespaces exist to prevent. Barnet is one of the 64.
+    render(<PlaceView detail={authority} englandAverage={61} neighbours={[]} />);
+    expect(screen.getByRole('link', { name: /^Primary schools in Barnet$/ }))
+      .toHaveAttribute('href', '/schools/authority/barnet/primary');
+    expect(screen.getByRole('link', { name: /^Secondary schools in Barnet$/ }))
+      .toHaveAttribute('href', '/schools/authority/barnet/secondary');
+  });
+
+  it('still uses the bare namespace for a town', () => {
+    render(<PlaceView detail={detail} englandAverage={61} neighbours={[]} />);
+    expect(screen.getByRole('link', { name: /^Primary schools in Brentwood$/ }))
+      .toHaveAttribute('href', '/schools/brentwood/primary');
+  });
+
+  it('offers no phase link when the place publishes none', () => {
+    // Outcodes are the case: no phase route exists for them, so the registry
+    // reports no phases and the nav does not render.
+    const outcode = { ...detail,
+      place: { ...detail.place, kind: 'outcode', slug: 'cm13', name: 'CM13',
+               phases: [] } };
+    render(<PlaceView detail={outcode} englandAverage={61} neighbours={[]} />);
+    expect(screen.queryByRole('navigation', { name: 'By phase' }))
+      .not.toBeInTheDocument();
+  });
+});
+
+describe('PlaceView unlinkable authorities', () => {
+  const withUnpublished: PlaceDetail = {
+    place: { kind: 'outcode', slug: 'tr21', name: 'TR21', count: 8,
+             parent_authority: 'Cornwall', phases: [],
+             authorities: [
+               { name: 'Cornwall', slug: 'cornwall', count: 6 },
+               { name: 'Isles Of Scilly', slug: null, count: 2 },
+             ] },
+    schools: [
+      { urn: 1, school_name: 'Alpha Primary', phase: 'Primary',
+        rwm_expected_pct: 82, attainment_8_score: null } as never,
+    ],
+    averages: { rwm_expected_pct: 63, attainment_8_score: null },
+  };
+
+  it('names an authority with no page without linking it', () => {
+    // City of London and the Isles of Scilly hold fewer schools than a page
+    // needs. Saying where the place is stays right; linking there would 404.
+    const { container } = render(<PlaceView detail={withUnpublished}
+      englandAverage={61} neighbours={[]} />);
+    expect(screen.getByRole('link', { name: 'Cornwall' })).toBeInTheDocument();
+    expect(screen.queryByRole('link', { name: 'Isles Of Scilly' }))
+      .not.toBeInTheDocument();
+    expect(container.querySelector('header p')?.textContent)
+      .toContain('Isles Of Scilly');
+  });
+});
+
+describe('PlaceView school attributes', () => {
+  /*
+   * The table shipped with one column of scores, which answers "how did they
+   * do" and nothing about whether the school is one a family could use. Age
+   * range, faith, nursery and constituency are the four facts a parent
+   * filters on before they look at a number at all.
+   */
+  const withAttributes: PlaceDetail = {
+    place: { kind: 'town', slug: 'chelmsford', name: 'Chelmsford', count: 3,
+             parent_authority: 'Essex', phases: ['primary', 'secondary'] },
+    schools: [
+      { urn: 1, school_name: 'Alpha Primary', phase: 'Primary',
+        rwm_expected_pct: 82, attainment_8_score: null,
+        age_range: '4-11', religious_denomination: 'Church of England',
+        nursery_provision: true,
+        parliamentary_constituency: 'Chelmsford' } as never,
+      { urn: 2, school_name: 'Beta High', phase: 'Secondary',
+        rwm_expected_pct: null, attainment_8_score: 47,
+        age_range: '11-16', religious_denomination: 'Does not apply',
+        nursery_provision: false,
+        parliamentary_constituency: 'Witham' } as never,
+    ],
+    averages: { rwm_expected_pct: 63, attainment_8_score: 45 },
+  };
+
+  function headings(container: HTMLElement, table = 0): string[] {
+    return Array.from(container.querySelectorAll('table')[table]
+      .querySelectorAll('thead th')).map((th) => th.textContent ?? '');
+  }
+
+  it('heads a primary table with all four attributes', () => {
+    const { container } = render(<PlaceView detail={withAttributes}
+      englandAverage={61} neighbours={[]} />);
+    expect(headings(container)).toEqual([
+      'School', 'Reading, writing & maths',
+      'Ages', 'Religious character', 'Nursery', 'Constituency',
+    ]);
+  });
+
+  it('omits nursery from a secondary table, where it does not apply', () => {
+    const { container } = render(<PlaceView detail={withAttributes}
+      englandAverage={61} neighbours={[]} />);
+    expect(headings(container, 1)).toEqual([
+      'School', 'Attainment 8', 'Ages', 'Religious character', 'Constituency',
+    ]);
+  });
+
+  it('keeps the measure beside the school name, where a phone can see it', () => {
+    // Six columns overflow a phone and .tableWrap turns that into a swipe.
+    // With the measure last, the one number the page exists for is the one
+    // scrolled off the screen.
+    const { container } = render(<PlaceView detail={withAttributes}
+      phase="primary" englandAverage={61} neighbours={[]} />);
+    expect(headings(container)[1]).toBe('Reading, writing & maths');
+  });
+
+  it('shows the age range without repeating the column heading', () => {
+    render(<PlaceView detail={withAttributes} englandAverage={61}
+      neighbours={[]} />);
+    expect(screen.getByText('4–11')).toBeInTheDocument();
+    expect(screen.queryByText('Ages 4–11')).not.toBeInTheDocument();
+  });
+
+  it('names the faith of a faith school', () => {
+    render(<PlaceView detail={withAttributes} englandAverage={61}
+      neighbours={[]} />);
+    expect(screen.getByText('Church of England')).toBeInTheDocument();
+  });
+
+  it('reads "Does not apply" as no religious character, not as a value', () => {
+    // GIAS spells the absence of a faith as "Does not apply", which is a
+    // database answer rather than an English one. The school page already
+    // suppresses it; the two must not disagree about the same school.
+    const { container } = render(<PlaceView detail={withAttributes}
+      englandAverage={61} neighbours={[]} />);
+    const secondary = container.querySelectorAll('table')[1]
+      .querySelectorAll('tbody td');
+    expect(secondary[3].textContent).toBe('—');
+    expect(screen.queryByText(/Does not apply/)).not.toBeInTheDocument();
+  });
+
+  it('marks a nursery as such and a school without one as not', () => {
+    const { container } = render(<PlaceView detail={withAttributes}
+      englandAverage={61} neighbours={[]} />);
+    const cells = container.querySelectorAll('table')[0]
+      .querySelectorAll('tbody td');
+    expect(cells[4].textContent).toBe('Yes');
+  });
+
+  it('names the constituency of each school', () => {
+    render(<PlaceView detail={withAttributes} englandAverage={61}
+      neighbours={[]} />);
+    expect(screen.getByText('Chelmsford', { selector: 'td' })).toBeInTheDocument();
+    expect(screen.getByText('Witham', { selector: 'td' })).toBeInTheDocument();
+  });
+
+  it('dashes an attribute the data does not carry', () => {
+    // nursery_provision and parliamentary_constituency are absent from marts
+    // the pipeline has not rebuilt, and the API degrades them to null rather
+    // than failing. A row must survive that.
+    const bare: PlaceDetail = {
+      ...withAttributes,
+      schools: [{ urn: 3, school_name: 'Gamma Primary', phase: 'Primary',
+                  rwm_expected_pct: 70 } as never],
+    };
+    const { container } = render(<PlaceView detail={bare} phase="primary"
+      englandAverage={61} neighbours={[]} />);
+    const cells = Array.from(container.querySelectorAll('tbody td'))
+      .map((td) => td.textContent);
+    expect(cells.slice(2)).toEqual(['—', '—', '—', '—']);
+  });
+
+  it('gives an all-through school its nursery under primary only', () => {
+    // All-through schools render in both groups. Nursery belongs to the
+    // primary reading of the same school, not the secondary one.
+    const allThrough: PlaceDetail = {
+      ...withAttributes,
+      schools: [{ urn: 4, school_name: 'Delta Academy', phase: 'All-through',
+                  rwm_expected_pct: 66, attainment_8_score: 51,
+                  age_range: '4-18', religious_denomination: 'None',
+                  nursery_provision: true,
+                  parliamentary_constituency: 'Chelmsford' } as never],
+    };
+    const { container } = render(<PlaceView detail={allThrough}
+      englandAverage={61} neighbours={[]} />);
+    const tables = container.querySelectorAll('table');
+    expect(tables[0].textContent).toContain('Yes');
+    expect(tables[1].textContent).not.toContain('Yes');
+  });
+});

nextjs-app/__tests__/components/Post16DestinationsSection.test.tsx

@@ -0,0 +1,44 @@
+import { render, screen } from '@testing-library/react';
+import { Post16DestinationsSection } from '@/components/school/Post16DestinationsSection';
+import type { DestinationPhase } from '@/lib/types';
+
+const phase: DestinationPhase = {
+  cohort_year: '2022/23',
+  groups: {
+    all: {
+      cohort: 96,
+      categories: [
+        { category: 'higher_education', pupils: 56, percentage: 58.3, status: 'published' },
+        { category: 'further_education', pupils: 12, percentage: 12.5, status: 'published' },
+        { category: 'apprenticeship', pupils: 9, percentage: 9.4, status: 'published' },
+        { category: 'employment', pupils: 13, percentage: 13.5, status: 'published' },
+        { category: 'not_sustained', pupils: 6, percentage: 6.3, status: 'published' },
+      ],
+    },
+  },
+};
+
+describe('Post16DestinationsSection', () => {
+  it('names the Year 13 cohort, not Year 11', () => {
+    const { container } = render(<Post16DestinationsSection destinations={phase} />);
+    expect(container.textContent).toMatch(/Year 13/);
+    expect(container.textContent).not.toMatch(/Year 11/);
+  });
+
+  it('reports higher education destinations', () => {
+    render(<Post16DestinationsSection destinations={phase} />);
+    expect(screen.getByText(/UK higher education/i)).toBeInTheDocument();
+  });
+
+  it('uses its own anchor so the nav does not collide with After Year 11', () => {
+    const { container } = render(<Post16DestinationsSection destinations={phase} />);
+    expect(container.querySelector('#post16-destinations')).toBeTruthy();
+    expect(container.querySelector('#destinations')).toBeNull();
+  });
+
+  it('renders nothing when no group carries categories', () => {
+    const empty: DestinationPhase = { cohort_year: '2022/23', groups: {} };
+    const { container } = render(<Post16DestinationsSection destinations={empty} />);
+    expect(container.firstChild).toBeNull();
+  });
+});

nextjs-app/__tests__/components/RouteTrail.test.tsx

@@ -0,0 +1,38 @@
+/**
+ * The trail has to be written by something, and it has to be written on every
+ * route — not only the ones that happen to track an event.
+ */
+import { render } from '@testing-library/react';
+
+const recordVisitedPath = jest.fn();
+let pathname = '/schools/brentwood';
+
+jest.mock('next/navigation', () => ({ usePathname: () => pathname }));
+jest.mock('@/lib/analytics', () => ({
+  recordVisitedPath: (p: string) => recordVisitedPath(p),
+}));
+
+// eslint-disable-next-line @typescript-eslint/no-var-requires
+const { RouteTrail } = require('@/components/RouteTrail');
+
+describe('RouteTrail', () => {
+  beforeEach(() => recordVisitedPath.mockClear());
+
+  it('records the page it is mounted on', () => {
+    render(<RouteTrail />);
+    expect(recordVisitedPath).toHaveBeenCalledWith('/schools/brentwood');
+  });
+
+  it('records each new route as the user moves through the app', () => {
+    const { rerender } = render(<RouteTrail />);
+    pathname = '/school/115429-brentwood-school';
+    rerender(<RouteTrail />);
+    expect(recordVisitedPath).toHaveBeenLastCalledWith(
+      '/school/115429-brentwood-school');
+  });
+
+  it('renders nothing, so it can sit anywhere in the layout', () => {
+    const { container } = render(<RouteTrail />);
+    expect(container).toBeEmptyDOMElement();
+  });
+});

nextjs-app/__tests__/components/SuggestList.test.tsx

@@ -0,0 +1,50 @@
+import { render, screen } from '@testing-library/react';
+import { SuggestList, suggestOptionId } from '@/components/SuggestList';
+
+const ROWS = [
+  { urn: 1, school_name: "St Mary's Primary", local_authority: 'Camden',
+    postcode: 'NW1 1AA', phase: 'Primary', school_type: 'Voluntary aided school' },
+  { urn: 2, school_name: "St Mary's Primary", local_authority: 'Barnet',
+    postcode: 'EN5 2AA', phase: 'Primary', school_type: 'Community school' },
+];
+
+describe('SuggestList', () => {
+  it('is a listbox of options', () => {
+    render(<SuggestList id="s" suggestions={ROWS} activeIndex={-1}
+      onPick={() => {}} onHover={() => {}} />);
+    expect(screen.getByRole('listbox')).toBeInTheDocument();
+    expect(screen.getAllByRole('option')).toHaveLength(2);
+  });
+
+  it('shows the local authority, which is what tells two schools apart', () => {
+    // Both rows are "St Mary's Primary". Without the authority the list is
+    // unusable for exactly the query autosuggest exists to serve.
+    render(<SuggestList id="s" suggestions={ROWS} activeIndex={-1}
+      onPick={() => {}} onHover={() => {}} />);
+    expect(screen.getByText('Camden')).toBeInTheDocument();
+    expect(screen.getByText('Barnet')).toBeInTheDocument();
+  });
+
+  it('marks only the active option selected', () => {
+    render(<SuggestList id="s" suggestions={ROWS} activeIndex={1}
+      onPick={() => {}} onHover={() => {}} />);
+    const options = screen.getAllByRole('option');
+    expect(options[0]).toHaveAttribute('aria-selected', 'false');
+    expect(options[1]).toHaveAttribute('aria-selected', 'true');
+  });
+
+  it('gives each option the id the input will point at', () => {
+    // aria-activedescendant on the input has to name a real element id, or
+    // a screen reader announces nothing as the user arrows through.
+    render(<SuggestList id="s" suggestions={ROWS} activeIndex={0}
+      onPick={() => {}} onHover={() => {}} />);
+    expect(screen.getAllByRole('option')[0]).toHaveAttribute(
+      'id', suggestOptionId('s', 0));
+  });
+
+  it('renders nothing when there is nothing to suggest', () => {
+    const { container } = render(<SuggestList id="s" suggestions={[]}
+      activeIndex={-1} onPick={() => {}} onHover={() => {}} />);
+    expect(container).toBeEmptyDOMElement();
+  });
+});

nextjs-app/__tests__/components/TrackPlaceView.test.tsx

@@ -0,0 +1,45 @@
+import { render } from '@testing-library/react';
+import { TrackPlaceView } from '@/components/places/TrackPlaceView';
+
+const trackMock = jest.fn();
+jest.mock('@/lib/analytics', () => ({
+  track: (...args: unknown[]) => trackMock(...args),
+  getNavigationSource: () => 'search',
+}));
+
+describe('TrackPlaceView', () => {
+  beforeEach(() => trackMock.mockClear());
+
+  it('reports which kind of location page was viewed', () => {
+    /*
+     * `kind` is the reason this event exists. Whether to keep investing in the
+     * location layer turns on which *sort* of page earns engagement — towns,
+     * authorities or postcode districts — and a bare pageview cannot say,
+     * because all four families share the /schools/ prefix.
+     */
+    render(<TrackPlaceView kind="authority" slug="kent" count={412} />);
+    expect(trackMock).toHaveBeenCalledWith('place_viewed', {
+      kind: 'authority', slug: 'kent', phase: 'all',
+      school_count: 412, from: 'search',
+    });
+  });
+
+  it('names the phase when the page is a phase variant', () => {
+    render(<TrackPlaceView kind="town" slug="brentwood" count={29} phase="primary" />);
+    expect(trackMock).toHaveBeenCalledWith('place_viewed',
+      expect.objectContaining({ phase: 'primary' }));
+  });
+
+  it('fires once, not once per render', () => {
+    const { rerender } = render(
+      <TrackPlaceView kind="town" slug="brentwood" count={29} />);
+    rerender(<TrackPlaceView kind="town" slug="brentwood" count={29} />);
+    expect(trackMock).toHaveBeenCalledTimes(1);
+  });
+
+  it('renders nothing', () => {
+    const { container } = render(
+      <TrackPlaceView kind="town" slug="brentwood" count={29} />);
+    expect(container).toBeEmptyDOMElement();
+  });
+});

nextjs-app/__tests__/components/darkThemeSafety.test.ts

@@ -0,0 +1,192 @@
+import fs from 'fs';
+import path from 'path';
+
+/**
+ * Guards against light-theme-only CSS.
+ *
+ * The site themes entirely through tokens redefined under
+ * `@media (prefers-color-scheme: dark)`. A hardcoded colour therefore does not
+ * fail loudly — it renders perfectly in the theme it was written for and
+ * quietly wrongly in the other, which nobody sees unless they happen to be in
+ * dark mode when they look.
+ *
+ * Both rules below are drawn from real defects in SchoolHeroMap.module.css,
+ * found by eye rather than by any test:
+ *
+ *  - the map's fade to the header ramped through hardcoded white and landed on
+ *    `var(--bg-card)`. Invisible in light; a bright band across the full width
+ *    of a near-black card in dark.
+ *  - the controls floating over the map paired a hardcoded white background
+ *    with `color: var(--text-primary)`, which resolves to #E9EEF0 in dark —
+ *    near-white text on a near-white button.
+ */
+
+const COMPONENTS = path.join(__dirname, '..', '..', 'components');
+
+function stylesheets(dir: string): string[] {
+  return fs.readdirSync(dir, { withFileTypes: true }).flatMap((entry) => {
+    const full = path.join(dir, entry.name);
+    if (entry.isDirectory()) return stylesheets(full);
+    return entry.name.endsWith('.module.css') ? [full] : [];
+  });
+}
+
+/** Innermost `selector { body }` pairs. Nested at-rules never match as rules,
+ *  because their body contains braces.
+ *
+ *  Comments are stripped before matching rather than after, so that the whole
+ *  selector survives. Taking only its last line — which is what stripping a
+ *  leading comment used to require — silently discarded every selector in a
+ *  grouped rule but the final one, and a safety guard that cannot see half its
+ *  input fails open. */
+function rules(css: string): Array<{ selector: string; body: string }> {
+  const bare = css.replace(/\/\*[\s\S]*?\*\//g, '');
+  return Array.from(bare.matchAll(/([^{}]+)\{([^{}]*)\}/g), (m) => ({
+    selector: m[1].trim().replace(/\s*\n\s*/g, ' '),
+    body: m[2],
+  }));
+}
+
+const HARDCODED_WHITE_BG = /background[^;]*(?:255,\s*255,\s*255|#fff\b|#ffffff\b)/i;
+const THEMED_COLOR = /(?:^|[^-])color:\s*var\(--/;
+
+const files = stylesheets(COMPONENTS);
+
+/** Component sources, for the third-party-surface rule below. */
+function sources(dir: string): string[] {
+  return fs.readdirSync(dir, { withFileTypes: true }).flatMap((entry) => {
+    const full = path.join(dir, entry.name);
+    if (entry.isDirectory()) return sources(full);
+    return entry.name.endsWith('.tsx') ? [full] : [];
+  });
+}
+
+describe('dark-theme safety', () => {
+  it('finds stylesheets to check', () => {
+    expect(files.length).toBeGreaterThan(0);
+  });
+
+  it('never pairs a hardcoded white background with a themed text colour', () => {
+    const offenders = files.flatMap((file) =>
+      rules(fs.readFileSync(file, 'utf8'))
+        .filter((r) => HARDCODED_WHITE_BG.test(r.body) && THEMED_COLOR.test(r.body))
+        .map((r) => `${path.relative(COMPONENTS, file)}  ${r.selector}`));
+
+    // Either the surface follows the theme and so should the text, or it does
+    // not and the text must be literal too. Mixing them is how near-white text
+    // ends up on a near-white button.
+    expect(offenders).toEqual([]);
+  });
+
+  it('never fades to a themed colour through a hardcoded one', () => {
+    const offenders = files.flatMap((file) =>
+      rules(fs.readFileSync(file, 'utf8'))
+        .filter((r) => /linear-gradient/.test(r.body)
+          && /var\(--bg-(card|primary|secondary)\)/.test(r.body)
+          && /255,\s*255,\s*255|#fff\b/i.test(r.body))
+        .map((r) => `${path.relative(COMPONENTS, file)}  ${r.selector}`));
+
+    // A gradient that lands on a token has to be made of that token, or the
+    // ramp and its destination disagree in one theme. Use the matching
+    // `--*-rgb` token for the transparent stops.
+    expect(offenders).toEqual([]);
+  });
+});
+
+/**
+ * The same defect one stylesheet further out.
+ *
+ * The rules above scan our own CSS modules. They cannot see a surface painted
+ * by a third-party sheet: leaflet.css hardcodes `background: white` on
+ * `.leaflet-popup-content-wrapper` and `.leaflet-popup-tip`, and
+ * LeafletMapInner builds its popup as an HTML string with inline
+ * `color: var(--text-primary)`. Neither half lives in a .module.css, so the
+ * module scan passed while dark mode rendered #E9EEF0 on #FFFFFF — 1.17:1,
+ * with the school name and the headline figure effectively invisible.
+ *
+ * globals.css already pulls the rest of Leaflet's chrome onto the tokens (the
+ * attribution bar, the zoom controls) for exactly this reason. The popup was
+ * simply missed.
+ */
+describe('third-party surfaces under themed text', () => {
+  const GLOBALS = path.join(__dirname, '..', '..', 'app', '(frontend)', 'globals.css');
+
+  /** Leaflet surfaces our own code writes token-coloured text onto. */
+  const LEAFLET_POPUP_SURFACES = [
+    '.leaflet-popup-content-wrapper',
+    '.leaflet-popup-tip',
+  ];
+
+  it('still finds a component painting themed text into a Leaflet popup', () => {
+    // Guards the rule below against passing vacuously if the popups are ever
+    // rewritten as React components rather than HTML strings.
+    const themed = sources(COMPONENTS).filter((file) => {
+      const src = fs.readFileSync(file, 'utf8');
+      return /bindPopup\(/.test(src) && /color:var\(--|color: var\(--/.test(src);
+    });
+
+    expect(themed.length).toBeGreaterThan(0);
+  });
+
+  it('themes the Leaflet popup surface, because the text on it is themed', () => {
+    const globals = rules(fs.readFileSync(GLOBALS, 'utf8'));
+
+    const unthemed = LEAFLET_POPUP_SURFACES.filter((surface) => {
+      const rule = globals.find((r) => r.selector.includes(surface));
+      return !rule || !/background[^;]*var\(--/.test(rule.body);
+    });
+
+    // Leaflet's white is not a colour this site owns. Either the surface
+    // follows the theme or the text on it must be literal — and the text is
+    // already themed.
+    expect(unthemed).toEqual([]);
+  });
+
+  it('never puts a literal white label on a themed fill', () => {
+    /*
+     * The mirror image of the module-CSS rule above, and the half of the popup
+     * that theming the card does not reach. "View Details" is
+     * `background:var(--status-above);color:white`; --status-above is #36743F
+     * in light but #7FCB8A in dark, so the label went from 5.63:1 to 1.94:1.
+     *
+     * --text-inverse is the token for ink on a saturated fill — #FFFFFF in
+     * light, #111A20 in dark — and the popup's Ofsted badge already uses it.
+     */
+    const offenders = sources(COMPONENTS).flatMap((file) => {
+      const src = fs.readFileSync(file, 'utf8');
+      return Array.from(
+        src.matchAll(/background:\s*var\(--[^;"']*;[^"']*?color:\s*(white|#fff\b|#ffffff\b)/gi),
+        () => path.relative(COMPONENTS, file));
+    });
+
+    expect(offenders).toEqual([]);
+  });
+});
+
+/**
+ * Destination measures add the first new colour family since the palette was
+ * set. The tokens have to exist in both blocks or the section renders one
+ * theme's fills on the other theme's ground — the exact failure the suite
+ * above exists to catch, but for tokens rather than literals.
+ */
+describe('destination tokens', () => {
+  const css = fs.readFileSync(
+    path.join(__dirname, '..', '..', 'app', '(frontend)', 'globals.css'), 'utf8');
+
+  const TOKENS = [
+    '--dest-sixthform', '--dest-sfcollege', '--dest-fecollege',
+    '--dest-apprentice', '--dest-employment', '--dest-none', '--dest-none-hatch',
+  ];
+
+  const DARK_AT = css.indexOf('@media (prefers-color-scheme: dark)');
+
+  it('defines every destination token in the light palette', () => {
+    const light = css.slice(0, DARK_AT);
+    expect(TOKENS.filter((t) => !light.includes(`${t}:`))).toEqual([]);
+  });
+
+  it('redefines every destination token for dark', () => {
+    const dark = css.slice(DARK_AT);
+    expect(TOKENS.filter((t) => !dark.includes(`${t}:`))).toEqual([]);
+  });
+});

nextjs-app/__tests__/components/filterBarScoping.test.ts

@@ -0,0 +1,108 @@
+import fs from 'fs';
+import path from 'path';
+
+/**
+ * The hero search and the results filter bar are the same component in two
+ * costumes. `.filterBar` is the card — background, border, shadow, padding —
+ * and `.heroMode` strips all of it so the search sits directly on the hero
+ * panel.
+ *
+ * Both selectors have specificity (0,1,0), so **source order decides**, and
+ * `.heroMode` only wins because it is declared immediately after. Any later
+ * bare `.filterBar` rule — which in practice means one inside a media query —
+ * silently wins instead, and the hero grows a card's padding back.
+ *
+ * That is exactly what happened: `@media (max-width: 768px) { .filterBar {
+ * padding: 0.875rem } }` re-added 14px in hero mode, indenting the search box,
+ * the hint and the location link 14px past the headline above them and costing
+ * the search field 28px of width on a 390px screen. The two rules directly
+ * below it in the same block were correctly written as
+ * `.filterBar:not(.heroMode)`; this one was missed, and nothing caught it
+ * because the result is a plausible-looking layout rather than a broken one.
+ */
+
+const CSS = path.join(__dirname, '..', '..', 'components', 'FilterBar.module.css');
+
+/** Properties `.heroMode` resets. A later bare `.filterBar` rule setting any
+ *  of these puts the card back on the hero. */
+const RESET_BY_HERO_MODE = [
+  'background', 'border', 'border-radius', 'box-shadow', 'padding',
+];
+
+/**
+ * Comments are stripped before anything is parsed.
+ *
+ * A `{` or `}` inside a comment would otherwise desynchronise the brace walk
+ * below and the rule regex alike, and the selector text captured for each rule
+ * would carry the preceding comment along with it.
+ */
+function withoutComments(css: string): string {
+  return css.replace(/\/\*[\s\S]*?\*\//g, '');
+}
+
+/**
+ * The individual selectors in a rule's prelude.
+ *
+ * Split on commas, because a selector list is a list: `.filterBar, .other { }`
+ * applies to `.filterBar` just as surely as `.filterBar { }` does, and an
+ * earlier version of this guard compared the whole prelude against the literal
+ * string '.filterBar' — so writing the regression as a comma list, or across
+ * two lines, would have walked straight past it.
+ */
+function selectorsOf(prelude: string): string[] {
+  return prelude.split(',').map((sel) => sel.trim().replace(/\s+/g, ' '))
+    .filter(Boolean);
+}
+
+function mediaQueryBodies(css: string): string[] {
+  const bodies: string[] = [];
+  const re = /@media[^{]*\{/g;
+  let m: RegExpExecArray | null;
+  while ((m = re.exec(css)) !== null) {
+    // Walk braces from the opening one to find this at-rule's whole body.
+    let depth = 1;
+    let i = m.index + m[0].length;
+    const start = i;
+    while (i < css.length && depth > 0) {
+      if (css[i] === '{') depth++;
+      else if (css[i] === '}') depth--;
+      i++;
+    }
+    bodies.push(css.slice(start, i - 1));
+  }
+  return bodies;
+}
+
+describe('FilterBar hero-mode scoping', () => {
+  const css = withoutComments(fs.readFileSync(CSS, 'utf8'));
+
+  it('confirms heroMode still resets the card, which is what makes this matter', () => {
+    const hero = css.match(/\.heroMode\s*\{([^}]*)\}/);
+    expect(hero).not.toBeNull();
+    expect(hero![1]).toMatch(/padding:\s*0/);
+  });
+
+  it('never re-applies card styling to the hero from inside a media query', () => {
+    const offenders: string[] = [];
+
+    for (const body of mediaQueryBodies(css)) {
+      for (const rule of body.matchAll(/([^{}]+)\{([^{}]*)\}/g)) {
+        // Only a *bare* .filterBar is dangerous, and it is dangerous wherever
+        // it appears in a selector list. Scoped variants
+        // (`.filterBar:not(.heroMode)`) and descendants are fine.
+        const selectors = selectorsOf(rule[1]);
+        if (!selectors.includes('.filterBar')) continue;
+
+        for (const prop of RESET_BY_HERO_MODE) {
+          if (new RegExp(`(^|[;\\s])${prop}\\s*:`).test(rule[2])) {
+            offenders.push(`${rule[1].trim()} sets ${prop}`);
+          }
+        }
+      }
+    }
+
+    // Fix by scoping the rule as `.filterBar:not(.heroMode)`, the way the
+    // neighbouring rules in the same block already are.
+    expect(offenders).toEqual([]);
+  });
+});

nextjs-app/__tests__/components/lastDistanceOffered.test.tsx

@@ -98,6 +98,17 @@ describe('secondary detail page', () => {
 
     expect(screen.getByText(/has not published a cut-off distance/)).toBeInTheDocument();
   });
+
+  it('makes no claim about publication when the feature is switched off', () => {
+    // Absent, not null. The API omits the key entirely while the
+    // admission_distance flag is off, and "Islington has not published a
+    // cut-off distance" is then a statement about us, not about Islington —
+    // false wherever the authority does publish one.
+    renderSecondarySchoolDetail({ ...secondaryFixture, admissionDistance: undefined });
+
+    expect(screen.queryByText(/has not published a cut-off distance/)).not.toBeInTheDocument();
+    expect(screen.queryByText(/Contact the admissions authority/)).not.toBeInTheDocument();
+  });
 });
 
 // ── The Distance section ───────────────────────────────────────────────

nextjs-app/__tests__/hooks/useSchoolSuggest.test.tsx

@@ -0,0 +1,73 @@
+import { renderHook, act, waitFor } from '@testing-library/react';
+import { useSchoolSuggest } from '@/hooks/useSchoolSuggest';
+
+const realFetch = global.fetch;
+
+function mockFetch(rows: unknown[], delayMs = 0) {
+  global.fetch = jest.fn(async (_url: unknown, init?: { signal?: AbortSignal }) => {
+    if (delayMs) {
+      await new Promise((resolve, reject) => {
+        const t = setTimeout(resolve, delayMs);
+        init?.signal?.addEventListener('abort', () => {
+          clearTimeout(t);
+          reject(Object.assign(new Error('aborted'), { name: 'AbortError' }));
+        });
+      });
+    }
+    return { ok: true, json: async () => ({ suggestions: rows }) };
+  }) as unknown as typeof fetch;
+}
+
+const ROW = {
+  urn: 1, school_name: 'Brecknock Primary School', local_authority: 'Camden',
+  postcode: 'NW1 1AA', phase: 'Primary', school_type: 'Community school',
+};
+
+describe('useSchoolSuggest', () => {
+  beforeEach(() => { jest.useFakeTimers(); });
+  afterEach(() => { jest.useRealTimers(); global.fetch = realFetch; });
+
+  it('does not fetch below the minimum query length', () => {
+    mockFetch([ROW]);
+    renderHook(() => useSchoolSuggest('b', true));
+    act(() => { jest.advanceTimersByTime(500); });
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+
+  it('does not fetch at all when disabled', () => {
+    // The flag being off must mean no request, not a hidden dropdown.
+    mockFetch([ROW]);
+    renderHook(() => useSchoolSuggest('brecknock', false));
+    act(() => { jest.advanceTimersByTime(500); });
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+
+  it('debounces rather than firing per keystroke', () => {
+    mockFetch([ROW]);
+    const { rerender } = renderHook(
+      ({ q }) => useSchoolSuggest(q, true), { initialProps: { q: 'br' } });
+    rerender({ q: 'bre' });
+    rerender({ q: 'brec' });
+    act(() => { jest.advanceTimersByTime(199); });
+    expect(global.fetch).not.toHaveBeenCalled();
+    act(() => { jest.advanceTimersByTime(2); });
+    expect(global.fetch).toHaveBeenCalledTimes(1);
+  });
+
+  it('opens with results once they arrive', async () => {
+    mockFetch([ROW]);
+    const { result } = renderHook(() => useSchoolSuggest('brecknock', true));
+    act(() => { jest.advanceTimersByTime(200); });
+    await waitFor(() => expect(result.current.suggestions).toHaveLength(1));
+    expect(result.current.open).toBe(true);
+  });
+
+  it('close() hides the list without clearing the query', async () => {
+    mockFetch([ROW]);
+    const { result } = renderHook(() => useSchoolSuggest('brecknock', true));
+    act(() => { jest.advanceTimersByTime(200); });
+    await waitFor(() => expect(result.current.open).toBe(true));
+    act(() => { result.current.close(); });
+    expect(result.current.open).toBe(false);
+  });
+});

nextjs-app/__tests__/lib/analytics.test.ts

@@ -0,0 +1,158 @@
+import { getNavigationSource } from '@/lib/analytics';
+
+/** jsdom's document.referrer is read-only; redefining it is the way in. */
+function referrer(url: string) {
+  Object.defineProperty(document, 'referrer', { value: url, configurable: true });
+}
+
+const ORIGIN = 'http://localhost';
+
+describe('getNavigationSource', () => {
+  afterEach(() => referrer(''));
+
+  it('attributes a visit from a location page to the place layer', () => {
+    /*
+     * The one this was added for.
+     *
+     * W2 published ~3,900 location pages whose entire purpose is to funnel
+     * search traffic onto school pages. Before this case existed they fell
+     * through to 'direct' — so the location layer's contribution was not
+     * merely missing from the funnel, it was being counted in the bucket you
+     * read as "typed the URL". The measurement that decides whether W2 worked
+     * was confidently reporting the wrong answer.
+     */
+    referrer(`${ORIGIN}/schools/barnet`);
+    expect(getNavigationSource()).toBe('place');
+  });
+
+  it.each([
+    ['/schools/authority/kent', 'authority'],
+    ['/schools/near/sw11', 'outcode'],
+    ['/schools/brentwood/primary', 'phase variant'],
+  ])('covers %s (%s)', (path) => {
+    referrer(`${ORIGIN}${path}`);
+    expect(getNavigationSource()).toBe('place');
+  });
+
+  it('still calls a school page "detail", one character away', () => {
+    // /school/ and /schools/ differ by one letter and mean different things.
+    // A prefix test written in the wrong order silently merges them.
+    referrer(`${ORIGIN}/school/100010-brecknock-primary-school`);
+    expect(getNavigationSource()).toBe('detail');
+  });
+
+  it.each([
+    ['/', 'search'],
+    ['/rankings', 'rankings'],
+    ['/compare?urns=1,2', 'compare'],
+  ])('leaves %s attributed as %s', (path, expected) => {
+    referrer(`${ORIGIN}${path}`);
+    expect(getNavigationSource()).toBe(expected);
+  });
+
+  it('treats an external referrer as direct', () => {
+    // Umami records the real referrer on the pageview; this field is only
+    // about internal navigation.
+    referrer('https://www.google.com/search?q=schools+in+barnet');
+    expect(getNavigationSource()).toBe('direct');
+  });
+
+  it('treats no referrer as direct', () => {
+    referrer('');
+    expect(getNavigationSource()).toBe('direct');
+  });
+});
+
+/*
+ * The defect the existing suite could not see.
+ *
+ * Every test above sets document.referrer, which the browser writes only when
+ * a *document* loads. Every internal navigation in this app is an App Router
+ * soft navigation — history.pushState, no new document — so document.referrer
+ * keeps naming whatever opened the tab for the whole session. Verified on
+ * staging: /schools/brentwood → click a school → URL changes to /school/…
+ * and document.referrer is still "".
+ *
+ * So `from` reported 'direct' for essentially every in-app journey, and the
+ * suite passed because it only ever exercised the full-page-load path.
+ */
+function freshAnalytics() {
+  let mod!: typeof import('@/lib/analytics');
+  jest.isolateModules(() => {
+    mod = require('@/lib/analytics');
+  });
+  return mod;
+}
+
+function at(path: string) {
+  window.history.pushState({}, '', path);
+}
+
+describe('getNavigationSource across a soft navigation', () => {
+  afterEach(() => {
+    referrer('');
+    at('/');
+  });
+
+  it('attributes a school view to the place page the user actually came from', () => {
+    const { recordVisitedPath, getNavigationSource: source } = freshAnalytics();
+    at('/schools/brentwood');
+    recordVisitedPath('/schools/brentwood');
+
+    at('/school/115429-brentwood-school');
+    recordVisitedPath('/school/115429-brentwood-school');
+
+    expect(source()).toBe('place');
+  });
+
+  it('does not depend on whether the new path was recorded first', () => {
+    // The trail is written by a layout-level effect and read by a page-level
+    // one. React orders those by tree position, which is not a contract worth
+    // resting a measurement on, so the answer must be the same either way.
+    const { recordVisitedPath, getNavigationSource: source } = freshAnalytics();
+    recordVisitedPath('/rankings');
+    at('/school/115429-brentwood-school');
+
+    expect(source()).toBe('rankings');
+  });
+
+  it('names the previous page, not the current one, when both are schools', () => {
+    const { recordVisitedPath, getNavigationSource: source } = freshAnalytics();
+    at('/school/100010-brecknock-primary-school');
+    recordVisitedPath('/school/100010-brecknock-primary-school');
+
+    at('/school/115429-brentwood-school');
+    recordVisitedPath('/school/115429-brentwood-school');
+
+    expect(source()).toBe('detail');
+  });
+
+  it('looks past a return visit to the page the user came back from', () => {
+    const { recordVisitedPath, getNavigationSource: source } = freshAnalytics();
+    for (const p of ['/schools/brentwood', '/school/115429-brentwood-school',
+                     '/schools/brentwood']) {
+      at(p);
+      recordVisitedPath(p);
+    }
+    expect(source()).toBe('detail');
+  });
+
+  it('falls back to the referrer on a real document load, where it is true', () => {
+    // A fresh module is a fresh document: nothing has been recorded, and
+    // document.referrer is meaningful again.
+    const { getNavigationSource: source } = freshAnalytics();
+    at('/school/115429-brentwood-school');
+    referrer(`${ORIGIN}/schools/barnet`);
+
+    expect(source()).toBe('place');
+  });
+
+  it('still reads an arrival from outside as direct', () => {
+    const { recordVisitedPath, getNavigationSource: source } = freshAnalytics();
+    at('/schools/brentwood');
+    recordVisitedPath('/schools/brentwood');
+    referrer('https://www.google.com/search?q=schools+in+brentwood');
+
+    expect(source()).toBe('direct');
+  });
+});

nextjs-app/__tests__/lib/destinations.test.ts

@@ -0,0 +1,87 @@
+import {
+  canAggregate, aggregateCells,
+  canRenderBar, toBarSegments, CARD_GROUPS,
+  type DestinationCell, type DestinationGroup, type DestinationCategory,
+} from '@/lib/destinations';
+
+const pub = (category: DestinationCategory, pupils: number, cohort: number): DestinationCell => ({
+  category, pupils, percentage: (pupils / cohort) * 100, status: 'published',
+});
+const sup = (category: DestinationCategory): DestinationCell => ({
+  category, pupils: null, percentage: null, status: 'suppressed',
+});
+
+const fullGroup = (): DestinationGroup => ({
+  cohort: 180,
+  cells: [
+    pub('school_sixth_form', 75, 180), pub('sixth_form_college', 21, 180),
+    pub('further_education', 55, 180), pub('other_education', 6, 180),
+    pub('apprenticeship', 8, 180), pub('employment', 6, 180),
+    pub('not_sustained', 5, 180), pub('not_captured', 4, 180),
+  ],
+});
+
+describe('canAggregate — R2, computing from components', () => {
+  it('allows a sum when every component is published', () => {
+    expect(canAggregate([pub('apprenticeship', 8, 180), pub('employment', 6, 180)])).toBe(true);
+  });
+
+  it('refuses a sum when any component is suppressed', () => {
+    expect(canAggregate([pub('apprenticeship', 8, 180), sup('employment')])).toBe(false);
+  });
+
+  it('refuses a sum when every component is suppressed', () => {
+    expect(canAggregate([sup('apprenticeship'), sup('employment')])).toBe(false);
+  });
+});
+
+describe('aggregateCells', () => {
+  it('sums published cells and derives a percentage from the cohort', () => {
+    expect(aggregateCells([pub('apprenticeship', 8, 180), pub('employment', 6, 180)], 180))
+      .toEqual({ pupils: 14, percentage: (14 / 180) * 100 });
+  });
+
+  it('returns null rather than a partial sum when a component is suppressed', () => {
+    expect(aggregateCells([pub('apprenticeship', 8, 180), sup('employment')], 180)).toBeNull();
+  });
+});
+
+describe('canRenderBar — R1', () => {
+  it('allows a bar when the whole group is published', () => {
+    expect(canRenderBar(fullGroup())).toBe(true);
+  });
+
+  it('refuses a bar when a single category is suppressed', () => {
+    const g = fullGroup();
+    g.cells[1] = sup('sixth_form_college');
+    expect(canRenderBar(g)).toBe(false);
+  });
+});
+
+describe('toBarSegments', () => {
+  it('derives widths from counts, not from rounded percentages', () => {
+    const segs = toBarSegments(fullGroup());
+    expect(segs).toHaveLength(8);
+    expect(segs[0].widthPct).toBeCloseTo((75 / 180) * 100, 10);
+    expect(segs.reduce((a, s) => a + s.widthPct, 0)).toBeCloseTo(100, 6);
+  });
+
+  it('throws rather than silently leaving a gap when the group is suppressed', () => {
+    const g = fullGroup();
+    g.cells[1] = sup('sixth_form_college');
+    expect(() => toBarSegments(g)).toThrow(/suppressed/i);
+  });
+});
+
+describe('CARD_GROUPS', () => {
+  it('partitions every destination category exactly once, plus the absence', () => {
+    const grouped = Object.values(CARD_GROUPS).flat();
+    expect(new Set(grouped).size).toBe(grouped.length);
+    expect(grouped).toEqual(expect.arrayContaining([
+      'school_sixth_form', 'sixth_form_college', 'further_education',
+      'other_education', 'apprenticeship', 'employment',
+    ]));
+    expect(grouped).not.toContain('not_sustained');
+    expect(grouped).not.toContain('not_captured');
+  });
+});

nextjs-app/__tests__/lib/flags.test.ts

@@ -0,0 +1,58 @@
+import { getFlags, FLAGS_REVALIDATE } from '@/lib/flags';
+
+// jsdom provides no global fetch, so there is nothing for jest.spyOn to attach
+// to — assign it and restore the original afterwards. This is the first test
+// here to mock fetch; later ones should follow this shape.
+const realFetch = global.fetch;
+
+function mockFetch(impl: () => Promise<unknown>) {
+  global.fetch = jest.fn(impl) as unknown as typeof fetch;
+}
+
+describe('getFlags', () => {
+  afterEach(() => { global.fetch = realFetch; });
+
+  it('returns the flags the API reports', async () => {
+    mockFetch(async () => ({
+      ok: true,
+      json: async () => ({ admission_distance: true }),
+    }));
+    await expect(getFlags()).resolves.toEqual({ admission_distance: true });
+  });
+
+  it('returns no flags rather than throwing when the API is down', async () => {
+    // A page that cannot read flags must render everything dark, not 500.
+    // Fail-closed is the same direction as the backend's default.
+    mockFetch(async () => { throw new Error('ECONNREFUSED'); });
+    await expect(getFlags()).resolves.toEqual({});
+  });
+
+  it('returns no flags rather than throwing on a non-200', async () => {
+    mockFetch(async () => ({ ok: false, status: 503 }));
+    await expect(getFlags()).resolves.toEqual({});
+  });
+
+  /*
+   * Reading a flag pins the calling route's ISR floor: Next uses the LOWEST
+   * revalidate among a route's fetches for the whole route. That is why the
+   * revalidate is an argument rather than the constant.
+   *
+   * Every SEO route here declares `revalidate = 604800`. A gate that read
+   * flags at the 300s default would drop the whole school and place corpus
+   * from a weekly cache to a 5-minute one, which is a large origin-load
+   * regression to pay for a feature flag.
+   */
+  it('reads at the 300s floor by default', async () => {
+    mockFetch(async () => ({ ok: true, json: async () => ({}) }));
+    await getFlags();
+    expect((global.fetch as jest.Mock).mock.calls[0][1])
+      .toEqual({ next: { revalidate: FLAGS_REVALIDATE } });
+  });
+
+  it('lets a caller pass its own route floor instead', async () => {
+    mockFetch(async () => ({ ok: true, json: async () => ({}) }));
+    await getFlags(604800);
+    expect((global.fetch as jest.Mock).mock.calls[0][1])
+      .toEqual({ next: { revalidate: 604800 } });
+  });
+});

nextjs-app/__tests__/lib/schoolSections.destinations.test.ts

@@ -0,0 +1,83 @@
+import { computeSecondaryFlags, buildSecondaryNavItems } from '@/lib/schoolSections';
+import type { School, SchoolDestinations } from '@/lib/types';
+
+const schoolInfo = {
+  urn: 137083, school_name: 'Northbrook Academy', phase: 'Secondary',
+  has_sixth_form: true,
+} as unknown as School;
+
+const base = { schoolInfo, yearlyData: [], deprivation: null, finance: null };
+
+const phase = (categories = 1) => ({
+  cohort_year: '2022/23',
+  groups: {
+    all: {
+      cohort: 180,
+      categories: Array.from({ length: categories }, () => ({
+        category: 'school_sixth_form' as const,
+        pupils: 75, percentage: 41.7, status: 'published' as const,
+      })),
+    },
+  },
+});
+
+const ks4Only: SchoolDestinations = { ks4: phase(), ks5: null };
+const both: SchoolDestinations = { ks4: phase(), ks5: phase() };
+
+describe('computeSecondaryFlags — destinations', () => {
+  it('flags KS4 destinations when the block carries categories', () => {
+    const flags = computeSecondaryFlags({ ...base, destinations: ks4Only });
+    expect(flags.hasKs4Destinations).toBe(true);
+    expect(flags.hasKs5Destinations).toBe(false);
+  });
+
+  it('flags both phases when both are present', () => {
+    const flags = computeSecondaryFlags({ ...base, destinations: both });
+    expect(flags.hasKs4Destinations).toBe(true);
+    expect(flags.hasKs5Destinations).toBe(true);
+  });
+
+  it('flags neither when the block is absent', () => {
+    const flags = computeSecondaryFlags({ ...base, destinations: null });
+    expect(flags.hasKs4Destinations).toBe(false);
+    expect(flags.hasKs5Destinations).toBe(false);
+  });
+
+  it('does not flag a phase whose groups carry no categories', () => {
+    const empty: SchoolDestinations = {
+      ks4: { cohort_year: '2022/23', groups: {} }, ks5: null,
+    };
+    expect(computeSecondaryFlags({ ...base, destinations: empty }).hasKs4Destinations)
+      .toBe(false);
+  });
+
+  it('does not flag a phase whose only group has an empty category list', () => {
+    const empty: SchoolDestinations = { ks4: phase(0), ks5: null };
+    expect(computeSecondaryFlags({ ...base, destinations: empty }).hasKs4Destinations)
+      .toBe(false);
+  });
+});
+
+describe('buildSecondaryNavItems — destinations', () => {
+  const navInput = {
+    ofsted: null, admissions: null, admissionDistance: null,
+    hasLocation: false, yearlyDataLength: 0,
+  };
+
+  it('adds both entries, after GCSEs', () => {
+    const flags = computeSecondaryFlags({ ...base, destinations: both });
+    const ids = buildSecondaryNavItems({ ...flags, hasResults: true }, navInput)
+      .map(i => i.id);
+    expect(ids).toContain('destinations');
+    expect(ids).toContain('post16-destinations');
+    expect(ids.indexOf('destinations')).toBeGreaterThan(ids.indexOf('gcse'));
+    expect(ids.indexOf('post16-destinations')).toBe(ids.indexOf('destinations') + 1);
+  });
+
+  it('adds no entry for a phase that will not render — the nav must not link to a missing anchor', () => {
+    const flags = computeSecondaryFlags({ ...base, destinations: null });
+    const ids = buildSecondaryNavItems(flags, navInput).map(i => i.id);
+    expect(ids).not.toContain('destinations');
+    expect(ids).not.toContain('post16-destinations');
+  });
+});

nextjs-app/__tests__/lib/utils.test.ts

@@ -13,6 +13,8 @@ import {
   metricKind,
   shortName,
   computeYBounds,
+  formatAgeRange,
+  formatAgeSpan,
 } from '@/lib/utils';
 
 describe('formatPercentage', () => {
@@ -320,3 +322,27 @@ describe('shortName', () => {
     expect(shortName('A'.repeat(30), 10)).toBe('AAAAAAAAA…');
   });
 });
+
+describe('formatAgeSpan', () => {
+  it('normalises a hyphenated range to an en dash, without a label', () => {
+    // The place table carries "Ages" in the column heading, so repeating it
+    // in every cell is noise. formatAgeRange keeps the label for the contexts
+    // that have no heading to hang it on.
+    expect(formatAgeSpan('4-11')).toBe('4–11');
+  });
+
+  it('leaves a range it does not recognise alone rather than mangling it', () => {
+    expect(formatAgeSpan('3-19 (SEN)')).toBe('3-19 (SEN)');
+  });
+
+  it('returns an empty string for a missing range', () => {
+    expect(formatAgeSpan(null)).toBe('');
+    expect(formatAgeSpan(undefined)).toBe('');
+  });
+});
+
+describe('formatAgeRange', () => {
+  it('keeps its label, so the two helpers stay distinguishable', () => {
+    expect(formatAgeRange('4-11')).toBe('Ages 4–11');
+  });
+});

nextjs-app/__tests__/payload/collections.test.ts

@@ -0,0 +1,70 @@
+/**
+ * Payload is ESM-only and next/jest will not transform it, so the collections
+ * cannot be imported and their sanitised config inspected here (see
+ * lib/payloadRoutes.ts for the full reasoning). These assert the source of the
+ * collection definitions instead — enough to catch the settings whose loss is
+ * silent, and cheap. Behaviour is proved by the e2e journeys against staging.
+ */
+import fs from 'fs';
+import path from 'path';
+
+const read = (file: string) =>
+  fs.readFileSync(path.join(__dirname, '..', '..', 'collections', file), 'utf8');
+
+const POSTS = read('Posts.ts');
+const MEDIA = read('Media.ts');
+const CONFIG = fs.readFileSync(
+  path.join(__dirname, '..', '..', 'payload.config.ts'),
+  'utf8',
+);
+
+describe('posts collection', () => {
+  it('supports drafts, so saving is not publishing', () => {
+    expect(POSTS).toMatch(/drafts:\s*true/);
+  });
+
+  it('has a unique, indexed slug for stable URLs', () => {
+    const slugField = POSTS.slice(POSTS.indexOf("name: 'slug'"));
+    expect(slugField).toMatch(/unique:\s*true/);
+    expect(slugField).toMatch(/index:\s*true/);
+  });
+
+  it('hides drafts from anonymous readers at the access layer', () => {
+    // Payload's docs are explicit: "The `draft` argument alone does not
+    // restrict documents with _status: 'draft' from being returned by the
+    // API." The blog pages' where-clause is not enforcement — a direct GET
+    // /cms-api/posts would return unpublished drafts to anyone. Access
+    // control returning a query constraint is the only thing that stops it.
+    expect(POSTS).toMatch(/_status:\s*\{\s*equals:\s*'published'\s*\}/);
+    expect(POSTS).toMatch(/if\s*\(req\.user\)\s*return true/);
+  });
+
+  it('revalidates the post page when a post changes or is deleted', () => {
+    // /blog/[slug] is ISR — generated on first request and cached — so an edit
+    // to an already-published post would otherwise not appear until the
+    // revalidate window expired, up to an hour of a writer concluding that
+    // saving is broken. The index and feeds are force-dynamic and need no hook.
+    expect(POSTS).toContain('afterChange');
+    expect(POSTS).toContain('afterDelete');
+    expect(POSTS).toMatch(/revalidatePath\(`\/blog\/\$\{[^}]+\}`\)/);
+  });
+});
+
+describe('media collection', () => {
+  it('writes uploads to the mounted volume, by absolute path', () => {
+    // Must match the payload_media mount in docker-compose.portainer.yml.
+    // Payload 3 requires staticDir to be absolute.
+    expect(MEDIA).toMatch(/staticDir:\s*'\/app\/media'/);
+  });
+
+  it('requires alt text on every upload', () => {
+    const altField = MEDIA.slice(MEDIA.indexOf("name: 'alt'"));
+    expect(altField).toMatch(/required:\s*true/);
+  });
+});
+
+describe('payload config', () => {
+  it('registers every collection', () => {
+    expect(CONFIG).toMatch(/collections:\s*\[Users,\s*Posts,\s*Media\]/);
+  });
+});

nextjs-app/__tests__/payload/import-map.test.ts

@@ -0,0 +1,67 @@
+/**
+ * The admin panel does not import field components directly. Payload sends the
+ * client a *path* for each one — a richText field's is
+ * `@payloadcms/richtext-lexical/rsc#RscEntryLexicalField` — and resolves it
+ * through this generated map. An entry that is missing from the map is not an
+ * error the panel reports: the field simply does not render.
+ *
+ * That failure is quietly awful, because `required: true` is enforced on the
+ * server regardless. A writer gets a new-post form with no Content editor and
+ * a save that refuses on a field they were never shown.
+ *
+ * The map is generated by `npx payload generate:importmap`, so it drifts every
+ * time a field or a lexical feature is added and nobody re-runs it. These
+ * assert the entries the current config needs.
+ */
+import fs from 'fs';
+import path from 'path';
+
+const MAP = fs.readFileSync(
+  path.join(__dirname, '..', '..', 'app', '(payload)', 'admin', 'importMap.js'),
+  'utf8',
+);
+const POSTS = fs.readFileSync(
+  path.join(__dirname, '..', '..', 'collections', 'Posts.ts'),
+  'utf8',
+);
+
+describe('admin import map', () => {
+  it('resolves the richText field, so Content renders in the editor', () => {
+    // Guarded because Posts.content is required: without this entry the field
+    // is invisible and the post is unsaveable.
+    expect(POSTS).toMatch(/type:\s*'richText'/);
+    expect(MAP).toContain('@payloadcms/richtext-lexical/rsc#RscEntryLexicalField');
+  });
+
+  it('resolves the richText cell, so the list view can render the column', () => {
+    expect(MAP).toContain('@payloadcms/richtext-lexical/rsc#RscEntryLexicalCell');
+  });
+
+  it('resolves the diff component, which the drafts UI needs', () => {
+    // versions.drafts is on, so the panel offers version comparison.
+    expect(POSTS).toMatch(/drafts:\s*true/);
+    expect(MAP).toContain('@payloadcms/richtext-lexical/rsc#LexicalDiffComponent');
+  });
+
+  it('resolves BlocksFeature, so the Callout block is insertable', () => {
+    expect(POSTS).toContain('BlocksFeature');
+    expect(MAP).toContain('@payloadcms/richtext-lexical/client#BlocksFeatureClient');
+  });
+
+  it('resolves the default toolbar features the editor is built with', () => {
+    // defaultFeatures is spread into the editor config; each one contributes a
+    // client component the toolbar cannot render without.
+    for (const feature of [
+      'BoldFeatureClient',
+      'ItalicFeatureClient',
+      'HeadingFeatureClient',
+      'LinkFeatureClient',
+      'UploadFeatureClient',
+      'UnorderedListFeatureClient',
+      'OrderedListFeatureClient',
+      'InlineToolbarFeatureClient',
+    ]) {
+      expect(MAP).toContain(`@payloadcms/richtext-lexical/client#${feature}`);
+    }
+  });
+});

nextjs-app/__tests__/payload/migrations.test.ts

@@ -0,0 +1,52 @@
+/**
+ * The generated migration is schema-qualified to "payload" throughout but does
+ * not create that schema — `schemaName` says where tables go, it does not
+ * create anything. On staging and production, which have never run it, the
+ * whole migration fails with `schema "payload" does not exist`.
+ *
+ * The CREATE SCHEMA is therefore hand-added, which makes it exactly the kind
+ * of edit a regeneration silently discards. This is the guard.
+ */
+import fs from 'fs';
+import path from 'path';
+
+const DIR = path.join(__dirname, '..', '..', 'migrations');
+
+function migrationFiles() {
+  return fs
+    .readdirSync(DIR)
+    .filter((f) => f.endsWith('.ts') && f !== 'index.ts');
+}
+
+describe('payload migrations', () => {
+  it('ships at least one migration, so a container has tables to find', () => {
+    expect(migrationFiles().length).toBeGreaterThan(0);
+  });
+
+  it('creates the payload schema before creating anything in it', () => {
+    const initial = migrationFiles().find((f) => f.includes('initial'))!;
+    const sql = fs.readFileSync(path.join(DIR, initial), 'utf8');
+
+    expect(sql).toMatch(/CREATE SCHEMA IF NOT EXISTS "payload"/);
+
+    // Ordering matters: the schema must be created before the first object
+    // that lives in it, or the migration fails on its first statement.
+    expect(sql.indexOf('CREATE SCHEMA IF NOT EXISTS "payload"'))
+      .toBeLessThan(sql.indexOf('CREATE TABLE "payload"'));
+  });
+
+  it('creates the tables the app queries on boot', () => {
+    const initial = migrationFiles().find((f) => f.includes('initial'))!;
+    const sql = fs.readFileSync(path.join(DIR, initial), 'utf8');
+    for (const table of ['users', 'posts', '_posts_v', 'media', 'payload_migrations']) {
+      expect(sql).toContain(`CREATE TABLE "payload"."${table}"`);
+    }
+  });
+
+  it('is wired into the adapter, so it runs on server init', () => {
+    const config = fs.readFileSync(
+      path.join(__dirname, '..', '..', 'payload.config.ts'), 'utf8',
+    );
+    expect(config).toMatch(/prodMigrations:\s*migrations/);
+  });
+});

nextjs-app/__tests__/payload/routes.test.ts

@@ -0,0 +1,45 @@
+/**
+ * Guards the one thing about Payload's mounting that fails silently.
+ *
+ * payload.config.ts itself cannot be imported here — Payload is ESM-only and
+ * next/jest will not transform it — so this asserts the shared constants and
+ * that the config actually wires them in, by reading its source. The live
+ * proof that /api still reaches FastAPI is the e2e journeys, which call
+ * /api/schools against the running app.
+ */
+import fs from 'fs';
+import path from 'path';
+import { PAYLOAD_API_ROUTE, PAYLOAD_ADMIN_ROUTE } from '@/lib/payloadRoutes';
+
+const CONFIG = fs.readFileSync(
+  path.join(__dirname, '..', '..', 'payload.config.ts'),
+  'utf8',
+);
+
+describe('payload mount points', () => {
+  it('serves the CMS API from /cms-api, never /api', () => {
+    // /api is the FastAPI proxy's catch-all. Payload's default would be
+    // swallowed by it and forwarded to the backend, silently.
+    expect(PAYLOAD_API_ROUTE).toBe('/cms-api');
+    expect(PAYLOAD_API_ROUTE).not.toBe('/api');
+  });
+
+  it('serves the admin panel from /admin', () => {
+    expect(PAYLOAD_ADMIN_ROUTE).toBe('/admin');
+  });
+
+  it('wires both constants into the Payload config', () => {
+    expect(CONFIG).toContain('PAYLOAD_API_ROUTE');
+    expect(CONFIG).toContain('PAYLOAD_ADMIN_ROUTE');
+  });
+
+  it('never hardcodes a routes block that could drift from the constants', () => {
+    expect(CONFIG).not.toMatch(/routes:\s*\{[^}]*api:\s*['"]/);
+  });
+
+  it('isolates CMS tables in their own postgres schema', () => {
+    // Blog content must sit outside `public`, where the app tables, Airflow's
+    // metadata and scripts/migrate_csv_to_db.py --drop all live.
+    expect(CONFIG).toMatch(/schemaName:\s*['"]payload['"]/);
+  });
+});

nextjs-app/__tests__/support/renderSchoolDetail.tsx

@@ -21,7 +21,7 @@ import {
 import { nationalAveragesFixture } from './schoolFixtures';
 
 // The shell calls useComparison(), which throws outside the provider. In the
-// app this wrapper comes from app/layout.tsx.
+// app this wrapper comes from app/(frontend)/layout.tsx.
 function withProviders(ui: ReactNode) {
   return <ComparisonProvider>{ui}</ComparisonProvider>;
 }

nextjs-app/app/(frontend)/about/About.module.css

@@ -0,0 +1,82 @@
+.page {
+  max-width: 42rem;
+  margin: 0 auto;
+  padding: 2.5rem 1.25rem 4rem;
+}
+
+.header {
+  display: flex;
+  align-items: center;
+  gap: 1.25rem;
+  margin-bottom: 2rem;
+}
+
+.portrait {
+  border-radius: 50%;
+  border: 2px solid var(--border);
+  object-fit: cover;
+  flex-shrink: 0;
+}
+
+.kicker {
+  font-family: var(--font-ui);
+  font-size: 0.75rem;
+  font-weight: 600;
+  text-transform: uppercase;
+  letter-spacing: 0.06em;
+  color: var(--brand);
+  margin: 0 0 0.35rem;
+}
+
+.heading {
+  font-family: var(--font-display);
+  font-size: clamp(1.5rem, 4vw, 2rem);
+  font-weight: 700;
+  line-height: 1.2;
+  color: var(--text-primary);
+  margin: 0;
+}
+
+.subheading {
+  font-family: var(--font-display);
+  font-size: 1.15rem;
+  font-weight: 600;
+  color: var(--text-primary);
+  margin: 2.25rem 0 0.75rem;
+}
+
+.prose p {
+  font-family: var(--font-ui);
+  font-size: 1rem;
+  line-height: 1.7;
+  color: var(--text-secondary);
+  margin: 0 0 1.1rem;
+}
+
+/* The opening paragraph carries the page. Larger, and in the primary ink
+   rather than the secondary, so it reads as a voice rather than as body copy.
+
+   Must stay in the descendant form: `.prose p` scores (0,1,1) and would beat a
+   bare `.lede` at (0,1,0), so simplifying this selector silently reverts the
+   lede to ordinary body copy. */
+.prose .lede {
+  font-size: 1.125rem;
+  color: var(--text-primary);
+}
+
+.link {
+  color: var(--brand);
+  font-weight: 600;
+}
+
+.link:hover {
+  color: var(--brand-strong);
+}
+
+@media (max-width: 480px) {
+  .header {
+    flex-direction: column;
+    align-items: flex-start;
+    gap: 1rem;
+  }
+}

nextjs-app/app/(frontend)/about/page.tsx

@@ -0,0 +1,143 @@
+import type { Metadata } from 'next';
+import Image from 'next/image';
+import { notFound } from 'next/navigation';
+import { absoluteUrl } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
+import { personJsonLd, organizationJsonLd } from '@/lib/jsonld';
+import styles from './About.module.css';
+
+export const metadata: Metadata = {
+  title: 'About',
+  description:
+    'Who builds schoolcompare, why it exists, and where its numbers come from.',
+  alternates: { canonical: absoluteUrl('/about') },
+};
+
+/*
+ * Gated on about_page. The default 300s read is the right floor here: this
+ * page declares no revalidate of its own, so nothing is lost by it, and a flip
+ * lands within five minutes.
+ *
+ * notFound(), not a redirect: while the flag is dark this URL does not exist,
+ * and a 404 is what tells a crawler not to keep it.
+ */
+export default async function AboutPage() {
+  const flags = await getFlags();
+  if (flags.about_page !== true) notFound();
+
+  const jsonLd = {
+    '@context': 'https://schema.org',
+    '@graph': [personJsonLd(), organizationJsonLd()],
+  };
+
+  return (
+    <div className={styles.page}>
+      <script
+        type="application/ld+json"
+        dangerouslySetInnerHTML={{ __html: JSON.stringify(jsonLd) }}
+      />
+
+      <header className={styles.header}>
+        <Image
+          src="/brand/tudor.jpg"
+          alt="Tudor, who builds schoolcompare"
+          width={96}
+          height={96}
+          className={styles.portrait}
+          priority
+        />
+        <div>
+          <p className={styles.kicker}>Who&apos;s behind this</p>
+          <h1 className={styles.heading}>I&apos;m Tudor. I built this site.</h1>
+        </div>
+      </header>
+
+      <div className={styles.prose}>
+        <p className={styles.lede}>
+          I&apos;m a parent in south-west London. When we started looking at
+          primary schools, I found the information I needed was all published,
+          and almost impossible to hold in one place.
+        </p>
+
+        <p>
+          SATs results were in one government table. Ofsted judgements were in a
+          separate service, in a format that had just changed. Admissions
+          distances were buried in council PDFs, a different one per borough,
+          each with its own layout. I ended up building a spreadsheet, and then
+          I got tired of the spreadsheet.
+        </p>
+
+        <p>
+          So I built this instead. It pulls the official figures into one place
+          and puts them side by side, which is what I wanted and could not find.
+        </p>
+
+        <h2 className={styles.subheading}>I&apos;m not an education expert</h2>
+
+        <p>
+          I want to be straightforward about that. I&apos;m not a teacher, a
+          governor, or an education researcher. I have no qualification that
+          makes my opinion about a school worth more than yours.
+        </p>
+
+        <p>
+          What I do have is the problem itself. I&apos;m going through primary
+          admissions right now, and I work with data for a living. That
+          combination is enough to take published figures and present them
+          honestly. It is not enough to tell you which school is right for your
+          child, and this site never tries to.
+        </p>
+
+        <h2 className={styles.subheading}>Where the numbers come from</h2>
+
+        <p>
+          Everything here is official published data: Key Stage 2 and Key Stage
+          4 results and school characteristics from the Department for
+          Education, inspection outcomes from Ofsted, and admissions data from
+          local authorities. Nothing is estimated, modelled or filled in. Where
+          a figure is missing, the page says so rather than showing a guess.
+        </p>
+
+        <p>
+          This is an independent site. It is not affiliated with the Department
+          for Education or with Ofsted, and nobody pays to appear on it or to
+          rank higher.
+        </p>
+
+        <h2 className={styles.subheading}>What the data can&apos;t tell you</h2>
+
+        <p>
+          A school is not its results. The figures here describe one year group,
+          on a handful of days, measured in a way that suits national statistics
+          rather than your child. A small cohort makes percentages swing wildly.
+          In a class of thirty, one pupil is worth more than three points.
+          Results say nothing at all about whether a child will be happy
+          somewhere.
+        </p>
+
+        <p>
+          I try to build that honesty into the site rather than just say it
+          here. Special schools and pupil referral units are never compared
+          against a mainstream national average, because that comparison is
+          meaningless and makes good schools look like failing ones. Where a
+          number is unreliable, the aim is for the page to tell you before you
+          draw a conclusion from it.
+        </p>
+
+        <h2 className={styles.subheading}>If something&apos;s wrong</h2>
+
+        <p>
+          Tell me and I&apos;ll fix it. If a figure looks wrong, or a page gives
+          a misleading impression of a school, I genuinely want to know.
+          It&apos;s the fastest way this gets better.
+        </p>
+
+        <p>
+          <a href="mailto:contact@schoolcompare.co.uk" className={styles.link}>
+            contact@schoolcompare.co.uk
+          </a>
+        </p>
+      </div>
+    </div>
+  );
+}

nextjs-app/app/(frontend)/api/[...path]/route.ts

@@ -26,8 +26,27 @@ function backendBase(): string {
 const STRIPPED_RESPONSE_HEADERS = ['content-encoding', 'content-length', 'transfer-encoding', 'connection'];
 const METHODS_WITH_BODY = new Set(['POST', 'PUT', 'PATCH', 'DELETE']);
 
+/*
+ * API paths this public proxy must not forward.
+ *
+ * Matched on the first segment, exactly — a prefix match would take
+ * /api/flagship down with /api/flags.
+ *
+ * `flags` is here because GET /api/flags names every unreleased feature the
+ * codebase knows about, along with whether it is on. Publishing that defeats
+ * the point of shipping dark. Next reads it server-side via FASTAPI_URL, on
+ * the Docker network, which never transits this route.
+ *
+ * Anything else internal-only belongs here too.
+ */
+const INTERNAL_ONLY_SEGMENTS = new Set(['flags']);
+
 async function handler(req: NextRequest, ctx: { params: Promise<{ path: string[] }> }) {
   const { path } = await ctx.params;
+  if (INTERNAL_ONLY_SEGMENTS.has(path[0])) {
+    return NextResponse.json({ detail: 'Not Found' }, { status: 404 });
+  }
+
   const target = `${backendBase()}/${path.join('/')}${req.nextUrl.search}`;
 
   const headers = new Headers(req.headers);

nextjs-app/app/(frontend)/blog/Blog.module.css

@@ -0,0 +1,76 @@
+.page {
+  max-width: 42rem;
+  margin: 0 auto;
+  padding: 2.5rem 1.25rem 4rem;
+}
+
+.header { margin-bottom: 2.5rem; }
+
+.kicker {
+  font-family: var(--font-ui);
+  font-size: 0.75rem;
+  font-weight: 600;
+  text-transform: uppercase;
+  letter-spacing: 0.06em;
+  color: var(--brand);
+  margin: 0 0 0.35rem;
+}
+
+.heading {
+  font-family: var(--font-display);
+  font-size: clamp(1.5rem, 4vw, 2rem);
+  font-weight: 700;
+  line-height: 1.2;
+  color: var(--text-primary);
+  margin: 0 0 0.75rem;
+}
+
+.standfirst {
+  font-family: var(--font-ui);
+  font-size: 1.05rem;
+  line-height: 1.65;
+  color: var(--text-secondary);
+  margin: 0;
+}
+
+.list { list-style: none; padding: 0; margin: 0; }
+
+.item {
+  padding: 1.5rem 0;
+  border-top: 1px solid var(--border);
+}
+
+.date {
+  font-family: var(--font-ui);
+  font-size: 0.8rem;
+  color: var(--text-muted);
+  /* Inter's tabular numerals keep a column of dates aligned. */
+  font-variant-numeric: tabular-nums;
+}
+
+.itemTitle {
+  font-family: var(--font-display);
+  font-size: 1.25rem;
+  font-weight: 600;
+  line-height: 1.3;
+  margin: 0.35rem 0 0.5rem;
+}
+
+.itemLink { color: var(--text-primary); text-decoration: none; }
+.itemLink:hover { color: var(--brand); }
+
+.excerpt {
+  font-family: var(--font-ui);
+  font-size: 0.95rem;
+  line-height: 1.65;
+  color: var(--text-secondary);
+  margin: 0;
+}
+
+.empty {
+  font-family: var(--font-ui);
+  color: var(--text-muted);
+}
+
+.link { color: var(--brand); font-weight: 600; }
+.link:hover { color: var(--brand-strong); }

nextjs-app/app/(frontend)/blog/[slug]/Post.module.css

@@ -0,0 +1,87 @@
+.page {
+  max-width: 42rem;
+  margin: 0 auto;
+  padding: 2.5rem 1.25rem 4rem;
+}
+
+.crumb {
+  font-family: var(--font-ui);
+  font-size: 0.85rem;
+  margin-bottom: 1.25rem;
+}
+
+.heading {
+  font-family: var(--font-display);
+  font-size: clamp(1.6rem, 5vw, 2.25rem);
+  font-weight: 700;
+  line-height: 1.2;
+  color: var(--text-primary);
+  margin: 0 0 0.75rem;
+}
+
+.byline {
+  font-family: var(--font-ui);
+  font-size: 0.9rem;
+  color: var(--text-muted);
+  margin: 0 0 2rem;
+}
+
+.hero {
+  width: 100%;
+  height: auto;
+  border-radius: 10px;
+  border: 1px solid var(--border);
+  margin-bottom: 2rem;
+}
+
+/* Rich-text output: the editor emits plain elements, so these are styled by
+   descendant selector rather than by class. */
+.prose p {
+  font-family: var(--font-ui);
+  font-size: 1rem;
+  line-height: 1.7;
+  color: var(--text-secondary);
+  margin: 0 0 1.1rem;
+}
+
+.prose h2 {
+  font-family: var(--font-display);
+  font-size: 1.25rem;
+  font-weight: 600;
+  color: var(--text-primary);
+  margin: 2.25rem 0 0.75rem;
+}
+
+.prose h3 {
+  font-family: var(--font-display);
+  font-size: 1.05rem;
+  font-weight: 600;
+  color: var(--text-primary);
+  margin: 1.75rem 0 0.6rem;
+}
+
+.prose ul,
+.prose ol {
+  font-family: var(--font-ui);
+  font-size: 1rem;
+  line-height: 1.7;
+  color: var(--text-secondary);
+  padding-left: 1.35rem;
+  margin: 0 0 1.1rem;
+}
+
+.prose li { margin-bottom: 0.4rem; }
+
+.prose a { color: var(--brand); font-weight: 500; }
+.prose a:hover { color: var(--brand-strong); }
+
+.prose blockquote {
+  border-left: 3px solid var(--border-strong);
+  padding-left: 1rem;
+  margin: 1.5rem 0;
+  color: var(--text-muted);
+  font-style: italic;
+}
+
+.link { color: var(--brand); font-weight: 600; }
+.link:hover { color: var(--brand-strong); }

nextjs-app/app/(frontend)/blog/[slug]/page.tsx

@@ -0,0 +1,194 @@
+import { cache } from 'react';
+import type { Metadata } from 'next';
+import Link from 'next/link';
+import { notFound } from 'next/navigation';
+import { RichText } from '@payloadcms/richtext-lexical/react';
+import type { JSXConvertersFunction } from '@payloadcms/richtext-lexical/react';
+import { getCachedPayload } from '@/lib/payload';
+import type { Post, Media } from '@/payload-types';
+import { absoluteUrl } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
+import {
+  blogPostingJsonLd,
+  breadcrumbJsonLd,
+  personJsonLd,
+  organizationJsonLd,
+} from '@/lib/jsonld';
+import { CalloutBlock } from '@/components/blog/CalloutBlock';
+import styles from './Post.module.css';
+
+/*
+ * ISR. Unlike the index, this route has a dynamic param and no
+ * generateStaticParams, so there is nothing for the build to prerender: each
+ * post is generated on first request and cached until the collection's
+ * afterChange hook revalidates it. That hook is what makes an edit to an
+ * already-published post appear immediately.
+ */
+export const revalidate = 3600;
+
+/**
+ * heroImage is `number | Media | null`: an id when the query is shallow, the
+ * populated document at depth 1. Both pages query at depth 1, but narrowing
+ * rather than asserting keeps it correct if that ever changes.
+ */
+function heroOf(post: Post): Media | null {
+  return typeof post.heroImage === 'object' && post.heroImage !== null
+    ? post.heroImage
+    : null;
+}
+
+/**
+ * Spreads the default converters and adds the one custom block.
+ *
+ * Without the spread, every default node type — paragraphs, headings, links —
+ * loses its renderer and the post body comes out empty.
+ */
+const calloutConverters: JSXConvertersFunction = ({ defaultConverters }) => ({
+  ...defaultConverters,
+  blocks: {
+    // Annotated because the generic block converter cannot infer a custom
+    // block's field shape; String() guards the values regardless.
+    callout: ({ node }: { node: { fields: Record<string, unknown> } }) => (
+      <CalloutBlock
+        tone={String(node.fields.tone ?? 'caveat')}
+        body={String(node.fields.body ?? '')}
+      />
+    ),
+  },
+});
+
+/**
+ * Wrapped in React's cache() because Next calls generateMetadata and the page
+ * component separately for the same request — without it, every post view runs
+ * this query against Postgres twice. cache() dedupes within a single request
+ * only, so it never serves one visitor's request from another's.
+ */
+const findPost = cache(async (slug: string) => {
+  const payload = await getCachedPayload();
+  const { docs } = await payload.find({
+    collection: 'posts',
+    where: { slug: { equals: slug }, _status: { equals: 'published' } },
+    limit: 1,
+    depth: 1,
+  });
+  return docs[0] ?? null;
+});
+
+function summarise(post: Post) {
+  return {
+    title: post.title,
+    slug: post.slug,
+    excerpt: post.excerpt,
+    publishedAt: post.publishedAt,
+  };
+}
+
+export async function generateMetadata(
+  { params }: { params: Promise<{ slug: string }> },
+): Promise<Metadata> {
+  const { slug } = await params;
+  const post = await findPost(slug);
+  if (!post) return { title: 'Not found' };
+
+  const hero = heroOf(post);
+
+  return {
+    title: post.title,
+    description: post.excerpt,
+    alternates: { canonical: absoluteUrl(`/blog/${post.slug}`) },
+    openGraph: {
+      type: 'article',
+      title: post.title,
+      description: post.excerpt,
+      url: absoluteUrl(`/blog/${post.slug}`),
+      publishedTime: post.publishedAt,
+      // A post with a hero image shares that; one without falls through to the
+      // generated share card at app/opengraph-image.tsx.
+      ...(hero?.url ? { images: [{ url: hero.url }] } : {}),
+    },
+  };
+}
+
+export default async function PostPage(
+  { params }: { params: Promise<{ slug: string }> },
+) {
+  const { slug } = await params;
+
+  /*
+   * Flags read at this route's own declared floor, so gating costs it nothing.
+   * Checked before the post is fetched: a dark blog should not query Payload.
+   */
+  const flags = await getFlags(3600);
+  if (flags.blog !== true) notFound();
+  const namedAuthor = flags.about_page === true;
+
+  const post = await findPost(slug);
+  if (!post) notFound();
+
+  const summary = summarise(post);
+  const hero = heroOf(post);
+
+  const jsonLd = {
+    '@context': 'https://schema.org',
+    /*
+     * The Person entity is anchored at /about#tudor, so it is declared only
+     * when that page exists. Claiming an author whose URL 404s is a worse
+     * signal than attributing the post to the publisher.
+     */
+    '@graph': [
+      blogPostingJsonLd(summary, { namedAuthor }),
+      breadcrumbJsonLd(summary),
+      ...(namedAuthor ? [personJsonLd()] : []),
+      organizationJsonLd(),
+    ],
+  };
+
+  return (
+    <article className={styles.page}>
+      <script
+        type="application/ld+json"
+        dangerouslySetInnerHTML={{ __html: JSON.stringify(jsonLd) }}
+      />
+
+      <nav className={styles.crumb}>
+        <Link href="/blog" className={styles.link}>Blog</Link>
+      </nav>
+
+      <h1 className={styles.heading}>{summary.title}</h1>
+
+      <p className={styles.byline}>
+        {/* Unlinked while about_page is dark; the flags are independent. */}
+        By {namedAuthor
+          ? <Link href="/about" className={styles.link}>Tudor</Link>
+          : 'Tudor'}
+        {' · '}
+        <time dateTime={summary.publishedAt}>
+          {new Date(summary.publishedAt).toLocaleDateString('en-GB', {
+            day: 'numeric',
+            month: 'long',
+            year: 'numeric',
+          })}
+        </time>
+      </p>
+
+      {/*
+        A plain <img>, not next/image: Payload already generated the sized
+        derivatives on upload (Media's imageSizes), so routing it through the
+        optimizer would resize an image that is already the right size.
+      */}
+      {hero?.url && (
+        <img
+          className={styles.hero}
+          src={hero.url}
+          alt={hero.alt ?? ''}
+          width={hero.width ?? undefined}
+          height={hero.height ?? undefined}
+        />
+      )}
+
+      <div className={styles.prose}>
+        <RichText data={post.content} converters={calloutConverters} />
+      </div>
+    </article>
+  );
+}

nextjs-app/app/(frontend)/blog/page.tsx

@@ -0,0 +1,85 @@
+import type { Metadata } from 'next';
+import Link from 'next/link';
+import { notFound } from 'next/navigation';
+import { getCachedPayload } from '@/lib/payload';
+import { absoluteUrl } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
+import styles from './Blog.module.css';
+
+/*
+ * Dynamic, not ISR.
+ *
+ * This route has no dynamic params, so Next prerenders it at build time — and
+ * CI builds the image with no database reachable, which fails the build. It is
+ * a single indexed query against Postgres on the same Docker network, so
+ * rendering per request is cheap, and it means a newly published post appears
+ * here immediately rather than waiting on a revalidation.
+ */
+export const dynamic = 'force-dynamic';
+
+export const metadata: Metadata = {
+  title: 'Blog',
+  description:
+    'Notes on what school performance data shows, and what it does not.',
+  alternates: { canonical: absoluteUrl('/blog') },
+};
+
+function formatDate(value: string) {
+  return new Date(value).toLocaleDateString('en-GB', {
+    day: 'numeric',
+    month: 'long',
+    year: 'numeric',
+  });
+}
+
+export default async function BlogIndexPage() {
+  const flags = await getFlags();
+  if (flags.blog !== true) notFound();
+
+  const payload = await getCachedPayload();
+  const { docs } = await payload.find({
+    collection: 'posts',
+    where: { _status: { equals: 'published' } },
+    sort: '-publishedAt',
+    limit: 50,
+    depth: 0,
+  });
+
+  return (
+    <div className={styles.page}>
+      <header className={styles.header}>
+        <p className={styles.kicker}>Blog</p>
+        <h1 className={styles.heading}>Notes on the numbers</h1>
+        <p className={styles.standfirst}>
+          What school performance data shows, what it doesn&apos;t, and how to
+          read it without being misled. Written by{' '}
+          {/* Plain text when about_page is dark: the two flags are
+              independent, so this link would otherwise point at a 404. */}
+          {flags.about_page === true
+            ? <Link href="/about" className={styles.link}>Tudor</Link>
+            : 'Tudor'}.
+        </p>
+      </header>
+
+      {docs.length === 0 ? (
+        <p className={styles.empty}>No posts yet.</p>
+      ) : (
+        <ul className={styles.list}>
+          {docs.map((post) => (
+            <li key={post.id} className={styles.item}>
+              <time className={styles.date} dateTime={String(post.publishedAt)}>
+                {formatDate(String(post.publishedAt))}
+              </time>
+              <h2 className={styles.itemTitle}>
+                <Link href={`/blog/${post.slug}`} className={styles.itemLink}>
+                  {post.title}
+                </Link>
+              </h2>
+              <p className={styles.excerpt}>{post.excerpt}</p>
+            </li>
+          ))}
+        </ul>
+      )}
+    </div>
+  );
+}

nextjs-app/app/(frontend)/blog/rss.xml/route.ts

@@ -0,0 +1,58 @@
+import { getCachedPayload } from '@/lib/payload';
+import { absoluteUrl } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
+
+/*
+ * Dynamic, not ISR.
+ *
+ * This route has no dynamic params, so Next prerenders it at build time — and
+ * CI builds the image with no database reachable, which fails the build. It is
+ * a single indexed query against Postgres on the same Docker network, so
+ * rendering per request is cheap, and it means a newly published post appears
+ * here immediately rather than waiting on a revalidation.
+ */
+export const dynamic = 'force-dynamic';
+
+function escapeXml(value: string): string {
+  return value.replace(/[<>&'"]/g, (char) =>
+    ({ '<': '&lt;', '>': '&gt;', '&': '&amp;', "'": '&apos;', '"': '&quot;' }[char]!));
+}
+
+export async function GET() {
+  // A dark blog has no feed. 404 rather than an empty channel: an empty feed
+  // is a live feed with nothing in it, which a reader would keep polling.
+  const flags = await getFlags();
+  if (flags.blog !== true) return new Response('Not found', { status: 404 });
+
+  const payload = await getCachedPayload();
+  const { docs } = await payload.find({
+    collection: 'posts',
+    where: { _status: { equals: 'published' } },
+    sort: '-publishedAt',
+    limit: 50,
+    depth: 0,
+  });
+
+  const items = docs.map((post) => `
+    <item>
+      <title>${escapeXml(String(post.title))}</title>
+      <link>${absoluteUrl(`/blog/${post.slug}`)}</link>
+      <guid isPermaLink="true">${absoluteUrl(`/blog/${post.slug}`)}</guid>
+      <description>${escapeXml(String(post.excerpt))}</description>
+      <pubDate>${new Date(String(post.publishedAt)).toUTCString()}</pubDate>
+    </item>`).join('');
+
+  const xml = `<?xml version="1.0" encoding="UTF-8"?>
+<rss version="2.0">
+  <channel>
+    <title>schoolcompare blog</title>
+    <link>${absoluteUrl('/blog')}</link>
+    <description>What school performance data shows, and what it does not.</description>
+    <language>en-GB</language>${items}
+  </channel>
+</rss>`;
+
+  return new Response(xml, {
+    headers: { 'Content-Type': 'application/rss+xml; charset=utf-8' },
+  });
+}

nextjs-app/app/(frontend)/content-sitemap.xml/route.ts

@@ -0,0 +1,68 @@
+/*
+ * A second sitemap for the URLs Next owns.
+ *
+ * /sitemap.xml is proxied from FastAPI (app/(frontend)/sitemap.xml), which
+ * knows nothing about Payload — the backend and frontend ship as separate
+ * images. Rather than teach it, the Next-owned URLs get their own sitemap and
+ * robots.txt lists both.
+ */
+import { getCachedPayload } from '@/lib/payload';
+import { absoluteUrl } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
+
+/*
+ * Dynamic, not ISR.
+ *
+ * This route has no dynamic params, so Next prerenders it at build time — and
+ * CI builds the image with no database reachable, which fails the build. It is
+ * a single indexed query against Postgres on the same Docker network, so
+ * rendering per request is cheap, and it means a newly published post appears
+ * here immediately rather than waiting on a revalidation.
+ */
+export const dynamic = 'force-dynamic';
+
+export async function GET() {
+  /*
+   * A dark page must not be advertised. Submitting a URL that 404s is the one
+   * thing a sitemap is not allowed to do, so each entry is gated on the same
+   * flag that gates the page itself.
+   *
+   * With both flags dark this emits a valid, empty <urlset> rather than a 404:
+   * robots.txt names this sitemap unconditionally, and an empty sitemap is a
+   * well-formed statement that there is nothing here yet.
+   */
+  const flags = await getFlags();
+  const aboutEnabled = flags.about_page === true;
+  const blogEnabled = flags.blog === true;
+
+  // Only query Payload when the blog is actually being advertised.
+  const docs = blogEnabled
+    ? (await (await getCachedPayload()).find({
+      collection: 'posts',
+      where: { _status: { equals: 'published' } },
+      sort: '-publishedAt',
+      limit: 500,
+      depth: 0,
+    })).docs
+    : [];
+
+  const urls: Array<{ loc: string; lastmod: string | null }> = [
+    ...(aboutEnabled ? [{ loc: absoluteUrl('/about'), lastmod: null }] : []),
+    ...(blogEnabled ? [{ loc: absoluteUrl('/blog'), lastmod: null }] : []),
+    ...docs.map((post) => ({
+      loc: absoluteUrl(`/blog/${post.slug}`),
+      lastmod: new Date(String(post.updatedAt ?? post.publishedAt)).toISOString(),
+    })),
+  ];
+
+  const xml = `<?xml version="1.0" encoding="UTF-8"?>
+<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+${urls.map(({ loc, lastmod }) =>
+  `  <url><loc>${loc}</loc>${lastmod ? `<lastmod>${lastmod}</lastmod>` : ''}</url>`,
+).join('\n')}
+</urlset>`;
+
+  return new Response(xml, {
+    headers: { 'Content-Type': 'application/xml; charset=utf-8' },
+  });
+}

nextjs-app/app/(frontend)/globals.css

@@ -28,6 +28,9 @@
     --bg-primary: #FAFAF8;      /* Warm White */
     --bg-secondary: #F5EFE6;    /* Sand — hero panels, sunken rows */
     --bg-card: #FFFFFF;
+    /* For gradients that have to fade to the card colour. A hardcoded white
+       ramp reads as a bright band against a dark card. */
+    --bg-card-rgb: 255, 255, 255;
     --surface-inverse: #0F766E;
 
     /* ── Ink ────────────────────────────────────────────────────────── */
@@ -102,6 +105,23 @@
     --series-7: #0E7A86;
     --series-8: #8A4A6B;
 
+    /* ── Destination measures ───────────────────────────────────────────
+       Education is one hue in three steps (school-like -> college-like) so the
+       education destinations read as one family; apprenticeship and employment
+       are separate hues. The absence is neutral and HATCHED, never a colour:
+       "activity not captured" covers independent schools, moving abroad and
+       training DfE holds no data on, so rendering it as a bad outcome would be
+       a factual error. The hatch is also the secondary encoding that rescues
+       the neutral/blue pair, which separates at only dE 7.6 as flat fills.
+       Every other adjacent pair clears dE 10.9 under protanopia. */
+    --dest-sixthform: #0F766E;
+    --dest-sfcollege: #4A9E96;
+    --dest-fecollege: #7CBFB8;
+    --dest-apprentice: #806200;
+    --dest-employment: #2F6F8F;
+    --dest-none: #6B7580;
+    --dest-none-hatch: rgba(107, 117, 128, 0.34);
+
     /* ── Phase: category, desaturated so it stays under the status hues ── */
     --phase-primary: #0F766E;
     --phase-primary-bg: rgba(167, 215, 197, 0.40);
@@ -234,6 +254,7 @@
         --bg-primary: #111A20;
         --bg-secondary: #16222A;
         --bg-card: #18242C;
+        --bg-card-rgb: 24, 36, 44;
         --surface-inverse: #E9EEF0;
 
         --text-primary: #E9EEF0;
@@ -289,6 +310,17 @@
         --series-7: #6FD0DC;
         --series-8: #D99BB8;
 
+        /* Destinations. Not a naive inversion: the education ramp reverses
+           direction so its darkest step stays the one furthest from the
+           school, and each step is re-checked against the dark card. */
+        --dest-sixthform: #5FC7BB;
+        --dest-sfcollege: #3E9B92;
+        --dest-fecollege: #2A716B;
+        --dest-apprentice: #EFC658;
+        --dest-employment: #8FB4D9;
+        --dest-none: #8B9AA1;
+        --dest-none-hatch: rgba(139, 154, 161, 0.34);
+
         --phase-primary: #5FC7BB;
         --phase-primary-bg: rgba(95, 199, 187, 0.16);
         --phase-primary-text: #8ADACF;
@@ -584,6 +616,35 @@ html .leaflet-bar a:hover {
     color: var(--text-primary);
 }
 
+/*
+ * The popup, which leaflet.css paints `background: white; color: #333` on both
+ * the card and its tip. The content LeafletMapInner binds into it is themed —
+ * the school name and the headline figure are `var(--text-primary)` — so in
+ * dark mode that was #E9EEF0 on #FFFFFF, a contrast ratio of 1.17:1. The name
+ * and the number were the two least readable things on the page.
+ *
+ * Moving the surface onto --bg-card fixes every foreground at once rather than
+ * one at a time: the muted phase line goes 2.90:1 -> 5.45:1, the vs-national
+ * delta 1.94:1 -> 8.14:1, the Ofsted badge 1.74:1 -> 9.11:1. In light mode
+ * --bg-card is #FFFFFF, so the popup looks as it always did.
+ */
+html .leaflet-popup-content-wrapper,
+html .leaflet-popup-tip {
+    background: var(--bg-card);
+    color: var(--text-primary);
+}
+
+/* Leaflet's own selector is `.leaflet-container a.leaflet-popup-close-button`
+   at 0,2,1 — an `html` prefix alone would lose to it. */
+html .leaflet-container a.leaflet-popup-close-button {
+    color: var(--text-muted);
+}
+
+html .leaflet-container a.leaflet-popup-close-button:hover,
+html .leaflet-container a.leaflet-popup-close-button:focus {
+    color: var(--text-primary);
+}
+
 /* Main content column */
 .main {
     max-width: 1400px;

nextjs-app/app/(frontend)/layout.tsx

@@ -4,8 +4,10 @@ import Script from 'next/script';
 import { Navigation } from '@/components/Navigation';
 import { Footer } from '@/components/Footer';
 import { ComparisonToast } from '@/components/ComparisonToast';
+import { RouteTrail } from '@/components/RouteTrail';
 import { ComparisonProvider } from '@/context/ComparisonProvider';
 import { SITE_URL } from '@/lib/site';
+import { getFlags } from '@/lib/flags';
 import './globals.css';
 
 // Manrope carries headings and key messaging — the guideline's "friendly,
@@ -77,11 +79,28 @@ export const metadata: Metadata = {
   },
 };
 
-export default function RootLayout({
+/*
+ * The footer's About and Blog links are flagged, which makes this the one
+ * place on the site that reads a flag on every route.
+ *
+ * 604800 is deliberate and load-bearing: it is the revalidate every SEO route
+ * here already declares. Next pins a route to the LOWEST revalidate among its
+ * fetches, so reading flags at the 300s default would drop the whole school
+ * and place corpus from a weekly cache to a 5-minute one — a large origin-load
+ * regression to hide two footer links.
+ *
+ * The cost is latency in one direction only. The pages themselves read the
+ * same flags at their own floors and flip within minutes; the footer links
+ * follow within a week. Turning a feature on early therefore shows the page
+ * before its footer link, which is harmless. Turning one off leaves a link to
+ * a 404 until the cache turns over, so a rollback that matters wants a purge.
+ */
+export default async function RootLayout({
   children,
 }: Readonly<{
   children: React.ReactNode;
 }>) {
+  const flags = await getFlags(604800);
   return (
     // The font variable classes must sit on <html>, not <body>. globals.css
     // declares --font-display on :root as var(--font-manrope) and --font-ui as
@@ -114,6 +133,10 @@ export default function RootLayout({
         />
       </head>
       <body>
+        {/* Records every route so funnel attribution has a previous page to
+            name. document.referrer cannot: a soft navigation creates no
+            document, so the browser never updates it. */}
+        <RouteTrail />
         <ComparisonProvider>
           <a href="#main-content" className="skip-link">Skip to main content</a>
           <Navigation />
@@ -121,7 +144,10 @@ export default function RootLayout({
             {children}
           </main>
           <ComparisonToast />
-          <Footer />
+          <Footer
+            aboutEnabled={flags.about_page === true}
+            blogEnabled={flags.blog === true}
+          />
         </ComparisonProvider>
       </body>
     </html>

nextjs-app/app/(frontend)/page.tsx

@@ -8,6 +8,7 @@ import type { Metadata } from 'next';
 import { fetchSchools, fetchFilters, fetchDataInfo } from '@/lib/api';
 import { formatAcademicYear } from '@/lib/utils';
 import { HomeView } from '@/components/HomeView';
+import { getFlags } from '@/lib/flags';
 import { HowItWorksSection } from '@/components/HowItWorksSection';
 import { EditorialSection } from '@/components/EditorialSection';
 
@@ -63,6 +64,11 @@ export default async function HomePage({ searchParams }: HomePageProps) {
   // Await search params (Next.js 15 requirement)
   const params = await searchParams;
 
+  // Server-read: no flag value reaches the browser bundle. Threaded down to
+  // both FilterBar instances via HomeView.
+  const flags = await getFlags();
+  const autosuggest = flags.school_autosuggest === true;
+
   // Parse search params
   const page = parseInt(params.page || '1');
   const radius = params.radius ? parseFloat(params.radius) : undefined;
@@ -111,6 +117,7 @@ export default async function HomePage({ searchParams }: HomePageProps) {
     const years = dataInfo?.years_available ?? [];
     return (
       <HomeView
+        autosuggest={autosuggest}
         initialSchools={schoolsData}
         filters={resolvedFilters}
         totalSchools={total}
@@ -131,6 +138,7 @@ export default async function HomePage({ searchParams }: HomePageProps) {
     const emptyFilters = { local_authorities: [], school_types: [], years: [], phases: [], genders: [], admissions_policies: [] };
     return (
       <HomeView
+        autosuggest={autosuggest}
         initialSchools={{ schools: [], page: 1, page_size: 50, total: 0, total_pages: 0 }}
         filters={emptyFilters}
         totalSchools={null}

nextjs-app/app/(frontend)/school/[slug]/page.tsx

@@ -148,7 +148,7 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
     notFound();
   }
 
-  const { school_info, yearly_data, absence_data, ofsted, census, admissions, admissions_history, admission_distance, deprivation, finance } = data;
+  const { school_info, yearly_data, absence_data, ofsted, census, admissions, admissions_history, admission_distance, deprivation, finance, destinations } = data;
 
   // Redirect bare URN to canonical slug URL
   const canonicalSlug = schoolUrl(urn, school_info.school_name).replace('/school/', '');
@@ -171,6 +171,7 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
     schoolInfo: school_info, yearlyData: yearly_data,
     absenceData: absence_data, census: census ?? null,
     deprivation: deprivation ?? null, finance: finance ?? null,
+    destinations: destinations ?? null,
   };
   const primaryFlags = computeSchoolFlags(sectionInput);
   const secondaryFlags = computeSecondaryFlags(sectionInput);
@@ -232,10 +233,11 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
             census={census ?? null}
             admissions={admissions ?? null}
             admissionsHistory={admissions_history ?? []}
-            admissionDistance={admission_distance ?? null}
+            admissionDistance={admission_distance}
             deprivation={deprivation ?? null}
             finance={finance ?? null}
             nationalAvg={nationalAvg}
+            destinations={destinations ?? null}
             flags={secondaryFlags}
           />
         </SchoolDetailShell>

nextjs-app/app/(frontend)/schools/[place]/[phase]/page.tsx

@@ -37,10 +37,12 @@ export async function generateMetadata({ params }: Props): Promise<Metadata> {
   const word = phase === 'secondary' ? 'Secondary' : 'Primary';
   const { name } = detail.place;
   return {
-    title: { absolute: `${word} Schools in ${name} — Ranked | schoolcompare` },
+    // Not "Ranked": the table is alphabetical, so the word would be a claim
+    // the page does not keep.
+    title: { absolute: `${word} Schools in ${name} | schoolcompare` },
     description:
-      `Every ${phase} school in ${name} ranked by results, with Ofsted grades and `
-      + `the local average against England.`,
+      `Every ${phase} school in ${name}, with results, Ofsted grades and the local `
+      + `average against England.`,
     alternates: { canonical: absoluteUrl(`/schools/${slug}/${phase}`) },
   };
 }

nextjs-app/app/(frontend)/schools/[place]/page.tsx

@@ -58,8 +58,8 @@ export async function generateMetadata({ params }: Props): Promise<Metadata> {
     // place title read '... | schoolcompare | schoolcompare'.
     title: { absolute: `Schools in ${name} — Compare ${count} Schools | schoolcompare` },
     description:
-      `Every school in ${name} ranked by SATs and GCSE results, with Ofsted grades, `
-      + `the local average against England, and how close you had to live to get a place.`,
+      `Every school in ${name}, with SATs and GCSE results, Ofsted grades, the local `
+      + `average against England, and how close you had to live to get a place.`,
     alternates: { canonical: absoluteUrl(`/schools/${slug}`) },
   };
 }
@@ -74,9 +74,14 @@ export default async function PlacePage({ params }: Props) {
   // defers to its authority rather than publishing a thin page.
   if (detail.averages.rwm_expected_pct == null
       && detail.averages.attainment_8_score == null) {
-    if (detail.place.parent_authority) {
-      redirect(`/schools/authority/${authoritySlug(detail.place.parent_authority)}`);
-    }
+    // The API's own slug, which is null when that authority is itself under
+    // the threshold and has no page. Re-slugifying the name here would send
+    // the reader to a 404 instead of telling them the place has no page.
+    const target = detail.place.authorities?.[0]?.slug
+      ?? (detail.place.parent_authority
+        ? authoritySlug(detail.place.parent_authority)
+        : null);
+    if (target) redirect(`/schools/authority/${target}`);
     notFound();
   }
 

nextjs-app/app/(frontend)/schools/authority/[la]/[phase]/page.tsx

@@ -0,0 +1,65 @@
+/**
+ * Phase variants of an authority page.
+ *
+ * The spec called for these; the plan built the bare authority route and
+ * dropped them. Nothing caught it, because the sitemap is written from the
+ * place registry — which was right about them all along — while the routes
+ * were written by hand. 302 authority phase URLs were submitted to Google and
+ * every one 404'd, and every authority page linked to a phase page in the
+ * *town* namespace, which is a different set of schools entirely.
+ *
+ * "Primary schools in Kent" is the query these serve, and it is a real one:
+ * admissions are authority-run, so the authority is the unit a parent thinks
+ * in when they have not settled on a town.
+ */
+import { notFound } from 'next/navigation';
+import type { Metadata } from 'next';
+import { fetchPlace } from '@/lib/places';
+import { fetchNationalAverages } from '@/lib/api';
+import { PlaceView } from '@/components/places/PlaceView';
+import { absoluteUrl } from '@/lib/site';
+
+interface Props { params: Promise<{ la: string; phase: string }> }
+
+export const revalidate = 604800;
+export const dynamicParams = true;
+
+const PHASES = ['primary', 'secondary'] as const;
+type Phase = (typeof PHASES)[number];
+
+const isPhase = (v: string): v is Phase => (PHASES as readonly string[]).includes(v);
+
+export async function generateMetadata({ params }: Props): Promise<Metadata> {
+  const { la, phase } = await params;
+  if (!isPhase(phase)) return { title: 'Place Not Found' };
+  const detail = await fetchPlace('authority', la, phase);
+  if (!detail || detail.schools.length === 0) return { title: 'Place Not Found' };
+
+  const word = phase === 'secondary' ? 'Secondary' : 'Primary';
+  const { name } = detail.place;
+  return {
+    // "Local Authority" stays in the title for the same reason it is on the
+    // bare authority page: 67 town names collide with an authority name, and
+    // a reader landing on both needs to know which set each covers.
+    title: { absolute: `${word} Schools in ${name} — Local Authority | schoolcompare` },
+    description:
+      `Every ${phase} school in the ${name} local authority, with results, Ofsted `
+      + `grades and the authority average against England.`,
+    alternates: { canonical: absoluteUrl(`/schools/authority/${la}/${phase}`) },
+  };
+}
+
+export default async function AuthorityPhasePage({ params }: Props) {
+  const { la, phase } = await params;
+  if (!isPhase(phase)) notFound();
+  const detail = await fetchPlace('authority', la, phase);
+  if (!detail || detail.schools.length === 0) notFound();
+
+  const national = await fetchNationalAverages().catch(() => null);
+  const englandAverage = phase === 'secondary'
+    ? national?.secondary?.attainment_8_score ?? null
+    : national?.primary?.rwm_expected_pct ?? null;
+
+  return <PlaceView detail={detail} phase={phase}
+    englandAverage={englandAverage} neighbours={[]} />;
+}

nextjs-app/app/(frontend)/schools/authority/[la]/page.tsx

@@ -45,8 +45,8 @@ export async function generateMetadata({ params }: Props): Promise<Metadata> {
   return {
     title: { absolute: `Schools in ${name} — Local Authority | schoolcompare` },
     description:
-      `All ${count} schools in the ${name} local authority, ranked by SATs and GCSE `
-      + `results, with Ofsted grades and the authority average against England.`,
+      `All ${count} schools in the ${name} local authority, with SATs and GCSE results, `
+      + `Ofsted grades and the authority average against England.`,
     alternates: { canonical: absoluteUrl(`/schools/authority/${la}`) },
   };
 }

nextjs-app/app/(frontend)/schools/near/[outcode]/page.tsx

@@ -40,8 +40,8 @@ export async function generateMetadata({ params }: Props): Promise<Metadata> {
   return {
     title: { absolute: `Schools near ${name} | schoolcompare` },
     description:
-      `${count} schools in the ${name} postcode district, ranked by results, with `
-      + `Ofsted grades and how close you had to live to get a place.`,
+      `${count} schools in the ${name} postcode district, with results, Ofsted grades `
+      + `and how close you had to live to get a place.`,
     alternates: { canonical: absoluteUrl(`/schools/near/${outcode}`) },
   };
 }

nextjs-app/app/(payload)/admin/[[...segments]]/not-found.tsx

@@ -0,0 +1,16 @@
+import type { Metadata } from 'next';
+import config from '@payload-config';
+import { NotFoundPage, generatePageMetadata } from '@payloadcms/next/views';
+import { importMap } from '../importMap.js';
+
+type Args = {
+  params: Promise<{ segments: string[] }>;
+  searchParams: Promise<{ [key: string]: string | string[] }>;
+};
+
+export const generateMetadata = ({ params, searchParams }: Args): Promise<Metadata> =>
+  generatePageMetadata({ config, params, searchParams });
+
+export default function NotFound({ params, searchParams }: Args) {
+  return NotFoundPage({ config, importMap, params, searchParams });
+}

nextjs-app/app/(payload)/admin/[[...segments]]/page.tsx

@@ -0,0 +1,16 @@
+import type { Metadata } from 'next';
+import config from '@payload-config';
+import { RootPage, generatePageMetadata } from '@payloadcms/next/views';
+import { importMap } from '../importMap.js';
+
+type Args = {
+  params: Promise<{ segments: string[] }>;
+  searchParams: Promise<{ [key: string]: string | string[] }>;
+};
+
+export const generateMetadata = ({ params, searchParams }: Args): Promise<Metadata> =>
+  generatePageMetadata({ config, params, searchParams });
+
+export default function Page({ params, searchParams }: Args) {
+  return RootPage({ config, importMap, params, searchParams });
+}

nextjs-app/app/(payload)/admin/importMap.js

@@ -0,0 +1,54 @@
+import { RscEntryLexicalCell as RscEntryLexicalCell_44fe37237e0ebf4470c9990d8cb7b07e } from '@payloadcms/richtext-lexical/rsc'
+import { RscEntryLexicalField as RscEntryLexicalField_44fe37237e0ebf4470c9990d8cb7b07e } from '@payloadcms/richtext-lexical/rsc'
+import { LexicalDiffComponent as LexicalDiffComponent_44fe37237e0ebf4470c9990d8cb7b07e } from '@payloadcms/richtext-lexical/rsc'
+import { BlocksFeatureClient as BlocksFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { BoldFeatureClient as BoldFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { ItalicFeatureClient as ItalicFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { UnderlineFeatureClient as UnderlineFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { StrikethroughFeatureClient as StrikethroughFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { SubscriptFeatureClient as SubscriptFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { SuperscriptFeatureClient as SuperscriptFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { InlineCodeFeatureClient as InlineCodeFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { ParagraphFeatureClient as ParagraphFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { HeadingFeatureClient as HeadingFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { AlignFeatureClient as AlignFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { IndentFeatureClient as IndentFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { UnorderedListFeatureClient as UnorderedListFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { OrderedListFeatureClient as OrderedListFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { ChecklistFeatureClient as ChecklistFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { LinkFeatureClient as LinkFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { RelationshipFeatureClient as RelationshipFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { BlockquoteFeatureClient as BlockquoteFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { UploadFeatureClient as UploadFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { HorizontalRuleFeatureClient as HorizontalRuleFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { InlineToolbarFeatureClient as InlineToolbarFeatureClient_e70f5e05f09f93e00b997edb1ef0c864 } from '@payloadcms/richtext-lexical/client'
+import { CollectionCards as CollectionCards_f9c02e79a4aed9a3924487c0cd4cafb1 } from '@payloadcms/next/rsc'
+
+/** @type import('payload').ImportMap */
+export const importMap = {
+  "@payloadcms/richtext-lexical/rsc#RscEntryLexicalCell": RscEntryLexicalCell_44fe37237e0ebf4470c9990d8cb7b07e,
+  "@payloadcms/richtext-lexical/rsc#RscEntryLexicalField": RscEntryLexicalField_44fe37237e0ebf4470c9990d8cb7b07e,
+  "@payloadcms/richtext-lexical/rsc#LexicalDiffComponent": LexicalDiffComponent_44fe37237e0ebf4470c9990d8cb7b07e,
+  "@payloadcms/richtext-lexical/client#BlocksFeatureClient": BlocksFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#BoldFeatureClient": BoldFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#ItalicFeatureClient": ItalicFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#UnderlineFeatureClient": UnderlineFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#StrikethroughFeatureClient": StrikethroughFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#SubscriptFeatureClient": SubscriptFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#SuperscriptFeatureClient": SuperscriptFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#InlineCodeFeatureClient": InlineCodeFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#ParagraphFeatureClient": ParagraphFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#HeadingFeatureClient": HeadingFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#AlignFeatureClient": AlignFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#IndentFeatureClient": IndentFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#UnorderedListFeatureClient": UnorderedListFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#OrderedListFeatureClient": OrderedListFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#ChecklistFeatureClient": ChecklistFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#LinkFeatureClient": LinkFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#RelationshipFeatureClient": RelationshipFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#BlockquoteFeatureClient": BlockquoteFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#UploadFeatureClient": UploadFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#HorizontalRuleFeatureClient": HorizontalRuleFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/richtext-lexical/client#InlineToolbarFeatureClient": InlineToolbarFeatureClient_e70f5e05f09f93e00b997edb1ef0c864,
+  "@payloadcms/next/rsc#CollectionCards": CollectionCards_f9c02e79a4aed9a3924487c0cd4cafb1
+}

nextjs-app/app/(payload)/cms-api/[...slug]/route.ts

@@ -0,0 +1,20 @@
+/*
+ * Payload's REST API, mounted at /cms-api rather than /api.
+ * See lib/payloadRoutes.ts — /api is the FastAPI proxy's catch-all.
+ */
+import config from '@payload-config';
+import {
+  REST_DELETE,
+  REST_GET,
+  REST_OPTIONS,
+  REST_PATCH,
+  REST_POST,
+  REST_PUT,
+} from '@payloadcms/next/routes';
+
+export const GET = REST_GET(config);
+export const POST = REST_POST(config);
+export const DELETE = REST_DELETE(config);
+export const PATCH = REST_PATCH(config);
+export const PUT = REST_PUT(config);
+export const OPTIONS = REST_OPTIONS(config);

nextjs-app/app/(payload)/cms-api/graphql-playground/route.ts

@@ -0,0 +1,4 @@
+import config from '@payload-config';
+import { GRAPHQL_PLAYGROUND_GET } from '@payloadcms/next/routes';
+
+export const GET = GRAPHQL_PLAYGROUND_GET(config);

nextjs-app/app/(payload)/cms-api/graphql/route.ts

@@ -0,0 +1,5 @@
+import config from '@payload-config';
+import { GRAPHQL_POST, REST_OPTIONS } from '@payloadcms/next/routes';
+
+export const POST = GRAPHQL_POST(config);
+export const OPTIONS = REST_OPTIONS(config);

nextjs-app/app/(payload)/layout.tsx

@@ -0,0 +1,27 @@
+/**
+ * Root layout for the Payload admin panel.
+ *
+ * This is a SECOND root layout: it renders its own <html>/<body>, as does
+ * app/(frontend)/layout.tsx. Next permits that only while no app/layout.tsx
+ * exists — which is why the site's routes were moved into (frontend). Adding
+ * an app/layout.tsx would nest the admin panel inside the site's nav, footer
+ * and providers and emit nested <html>.
+ */
+import type { ServerFunctionClient } from 'payload';
+import config from '@payload-config';
+import { RootLayout, handleServerFunctions } from '@payloadcms/next/layouts';
+import { importMap } from './admin/importMap.js';
+import '@payloadcms/next/css';
+
+const serverFunction: ServerFunctionClient = async function (args) {
+  'use server';
+  return handleServerFunctions({ ...args, config, importMap });
+};
+
+export default function PayloadLayout({ children }: { children: React.ReactNode }) {
+  return (
+    <RootLayout config={config} importMap={importMap} serverFunction={serverFunction}>
+      {children}
+    </RootLayout>
+  );
+}

nextjs-app/app/robots.ts

@@ -12,9 +12,14 @@ export default function robots(): MetadataRoute.Robots {
       {
         userAgent: '*',
         allow: '/',
-        disallow: ['/api/', '/_next/'],
+        // /admin and /cms-api are also served X-Robots-Tag: noindex by
+        // next.config.mjs. A Disallow alone blocks crawling, not indexing.
+        disallow: ['/api/', '/_next/', '/admin/', '/cms-api/'],
       },
     ],
-    sitemap: absoluteUrl('/sitemap.xml'),
+    // Two sitemaps: /sitemap.xml is proxied from FastAPI and carries the
+    // school corpus; /content-sitemap.xml is Next-owned and carries /about
+    // and the blog. The backend knows nothing about Payload.
+    sitemap: [absoluteUrl('/sitemap.xml'), absoluteUrl('/content-sitemap.xml')],
   };
 }

nextjs-app/blocks/Callout.ts

@@ -0,0 +1,25 @@
+import type { Block } from 'payload';
+
+/**
+ * The house block: "what this number doesn't tell you".
+ *
+ * Blocks are the reason this site runs a CMS rather than flat files — a post
+ * can carry live product components, not screenshots of them. This is the
+ * first and simplest one; a live-chart block follows when a post needs it.
+ */
+export const Callout: Block = {
+  slug: 'callout',
+  labels: { singular: 'Callout', plural: 'Callouts' },
+  fields: [
+    {
+      name: 'tone',
+      type: 'select',
+      defaultValue: 'caveat',
+      options: [
+        { label: 'Caveat: what this does not show', value: 'caveat' },
+        { label: 'Note: useful aside', value: 'note' },
+      ],
+    },
+    { name: 'body', type: 'textarea', required: true },
+  ],
+};

nextjs-app/collections/Media.ts

@@ -0,0 +1,31 @@
+import type { CollectionConfig } from 'payload';
+
+/**
+ * Uploads land on a Docker named volume mounted at /app/media. The path is
+ * absolute because Payload 3 requires it, and it must match the payload_media
+ * mount in docker-compose.portainer.yml exactly — a mismatch writes into the
+ * container's own filesystem, where the next redeploy silently discards it.
+ */
+export const Media: CollectionConfig = {
+  slug: 'media',
+  access: { read: () => true },
+  upload: {
+    staticDir: '/app/media',
+    mimeTypes: ['image/*'],
+    imageSizes: [
+      { name: 'thumbnail', width: 400 },
+      { name: 'hero', width: 1200 },
+    ],
+    adminThumbnail: 'thumbnail',
+  },
+  fields: [
+    {
+      name: 'alt',
+      type: 'text',
+      required: true,
+      // Required rather than optional: a decorative-by-default image is an
+      // accessibility regression on a site parents use under time pressure.
+      admin: { description: 'Describe the image for screen readers.' },
+    },
+  ],
+};

nextjs-app/collections/Posts.ts

@@ -0,0 +1,99 @@
+import type { CollectionConfig } from 'payload';
+import { revalidatePath } from 'next/cache';
+import { lexicalEditor, BlocksFeature } from '@payloadcms/richtext-lexical';
+import { Callout } from '@/blocks/Callout';
+
+/**
+ * Drop the cached copy of a post page when it changes.
+ *
+ * Only the post page needs this. The blog index, the RSS feed and the content
+ * sitemap are force-dynamic — they have no dynamic params, so Next would
+ * prerender them at build time, where CI has no database — which means they
+ * already reflect a change on the next request.
+ *
+ * /blog/[slug] is ISR: generated on first request and cached, so without this
+ * an edit to an already-published post would not appear until the revalidate
+ * window expired — up to an hour of a writer concluding that saving is broken.
+ *
+ * Payload runs in the same process as Next, so this is a direct revalidatePath
+ * call: no webhook, no shared secret, no network hop to get wrong.
+ */
+function revalidatePost(slug: string) {
+  revalidatePath(`/blog/${slug}`);
+}
+
+export const Posts: CollectionConfig = {
+  slug: 'posts',
+  access: {
+    /*
+     * Drafts must be hidden here, not in the pages that query this collection.
+     *
+     * From Payload's own documentation: "The `draft` argument alone does not
+     * restrict documents with `_status: 'draft'` from being returned by the
+     * API." The blog index and post page both filter on `_status`, but that
+     * is a convenience, not a control — a direct GET /cms-api/posts would
+     * hand every unpublished draft to any visitor.
+     *
+     * Returning a query constraint rather than a boolean is the documented
+     * mechanism: Payload merges it into every read for an anonymous caller.
+     */
+    read: ({ req }) => {
+      if (req.user) return true;
+      return { _status: { equals: 'published' } };
+    },
+  },
+  admin: {
+    useAsTitle: 'title',
+    defaultColumns: ['title', 'publishedAt', '_status'],
+  },
+  versions: {
+    // Posts get written across several sittings and previewed before they go
+    // live. Without drafts, saving is publishing.
+    drafts: true,
+  },
+  hooks: {
+    afterChange: [({ doc }) => { revalidatePost(String(doc.slug)); }],
+    afterDelete: [({ doc }) => { revalidatePost(String(doc.slug)); }],
+  },
+  fields: [
+    { name: 'title', type: 'text', required: true },
+    {
+      name: 'slug',
+      type: 'text',
+      required: true,
+      unique: true,
+      index: true,
+      admin: {
+        position: 'sidebar',
+        description: 'The URL segment. Never change it after publishing.',
+      },
+    },
+    {
+      name: 'publishedAt',
+      type: 'date',
+      required: true,
+      admin: { position: 'sidebar', date: { pickerAppearance: 'dayOnly' } },
+    },
+    {
+      name: 'excerpt',
+      type: 'textarea',
+      required: true,
+      maxLength: 200,
+      admin: {
+        description: 'Shown on the index and used as the meta description.',
+      },
+    },
+    { name: 'heroImage', type: 'upload', relationTo: 'media' },
+    {
+      name: 'content',
+      type: 'richText',
+      required: true,
+      editor: lexicalEditor({
+        features: ({ defaultFeatures }) => [
+          ...defaultFeatures,
+          BlocksFeature({ blocks: [Callout] }),
+        ],
+      }),
+    },
+  ],
+};

nextjs-app/collections/Users.ts

@@ -0,0 +1,33 @@
+import type { CollectionConfig } from 'payload';
+
+/**
+ * The site's only authenticated surface. There is one account and no
+ * registration: `create` is closed to everyone, so the first user is seeded
+ * with `payload create-first-user` and no one can add another through the API.
+ */
+export const Users: CollectionConfig = {
+  slug: 'users',
+  auth: {
+    // Slows credential stuffing against a panel that is on the public
+    // internet. Five attempts, then a ten-minute lock.
+    maxLoginAttempts: 5,
+    lockTime: 10 * 60 * 1000,
+  },
+  access: {
+    create: () => false,
+    read: ({ req }) => Boolean(req.user),
+    update: ({ req }) => Boolean(req.user),
+    delete: () => false,
+  },
+  admin: { useAsTitle: 'email' },
+  fields: [
+    {
+      name: 'displayName',
+      type: 'text',
+      required: true,
+      // Rendered as the byline on every post. First name only — the site
+      // publishes no surname and no employer.
+      defaultValue: 'Tudor',
+    },
+  ],
+};

nextjs-app/components/FilterBar.module.css

@@ -48,6 +48,8 @@
   display: flex;
   align-items: center;
   gap: 0.5rem;
+  /* The suggestion dropdown is absolutely positioned against this box. */
+  position: relative;
 }
 
 /* The hero pill: hairline, soft corner, everything else sits inside it. */
@@ -411,7 +413,17 @@
 /* ── Narrow ───────────────────────────────────────────────────────── */
 
 @media (max-width: 768px) {
-  .filterBar {
+  /*
+   * Scoped, like the two rules below it.
+   *
+   * The results filter bar is a card — background, border, shadow — and needs
+   * inner padding. The hero's search is not a card: .heroMode zeroes the
+   * padding, border and background so the search sits directly on the panel.
+   * Unscoped, this rule put 14px back, which indented the search box, the hint
+   * and the location link 14px past the headline they sit under, and cost the
+   * search field 28px of width on a 390px screen.
+   */
+  .filterBar:not(.heroMode) {
     padding: 0.875rem;
   }
 
@@ -455,6 +467,14 @@
     align-items: flex-start;
   }
 
+  /* Optical alignment: the button's own 6px of padding is what makes its
+     label start further right than the hint above it, even once both boxes
+     share a left edge. Pulling the padding back off lines the text up while
+     keeping the tap target. */
+  .heroMode .nearMeBtn {
+    margin-left: -0.375rem;
+  }
+
   .geoError {
     text-align: left;
   }

nextjs-app/components/FilterBar.tsx

@@ -3,8 +3,11 @@
 import { useState, useCallback, useTransition, useRef, useEffect } from "react";
 import type { ReactNode } from "react";
 import { useRouter, useSearchParams, usePathname } from "next/navigation";
-import { isValidPostcode } from "@/lib/utils";
+import { isValidPostcode, schoolUrl } from "@/lib/utils";
 import { track } from "@/lib/analytics";
+import { useSchoolSuggest } from "@/hooks/useSchoolSuggest";
+import { SuggestList, suggestOptionId } from "./SuggestList";
+import type { Suggestion } from "@/lib/suggest";
 import type { Filters, ResultFilters } from "@/lib/types";
 import styles from "./FilterBar.module.css";
 
@@ -17,6 +20,8 @@ interface FilterBarProps {
   onNearMe?: () => void;
   geoState?: "idle" | "requesting" | "error";
   geoError?: string | null;
+  /** Server-read feature flag. Off means no listener, no fetch, no markup. */
+  autosuggest?: boolean;
 }
 
 /**
@@ -48,6 +53,7 @@ export function FilterBar({
   onNearMe,
   geoState = "idle",
   geoError,
+  autosuggest = false,
 }: FilterBarProps) {
   const router = useRouter();
   const pathname = usePathname();
@@ -62,6 +68,59 @@ export function FilterBar({
 
   const [omniValue, setOmniValue] = useState(initialOmniValue);
 
+  const suggestId = `school-suggest-${isHero ? "hero" : "bar"}`;
+
+  /*
+   * Suggestions answer typing, not the mere presence of a value.
+   *
+   * Without this the results-page bar reopened the dropdown over the results:
+   * after a search the input still holds the term, so on every render the
+   * query was >= 2 characters and the list opened again — on top of the very
+   * results the search had just produced, swallowing the click on the first
+   * one. The E2E gate caught it as "<li role=option> intercepts pointer
+   * events", but a reader would just have found the page unclickable.
+   */
+  const [hasTyped, setHasTyped] = useState(false);
+
+  // Suppressed once the value parses as a postcode: the box takes a school
+  // name OR a postcode, and suggesting schools during postcode entry fights
+  // the user rather than helping them.
+  const suggestEnabled = autosuggest && hasTyped && !isValidPostcode(omniValue);
+  const { suggestions, open, activeIndex, setActiveIndex, close } =
+    useSchoolSuggest(omniValue, suggestEnabled);
+
+  const pickSuggestion = (s: Suggestion) => {
+    setHasTyped(false);
+    close();
+    track('search_submitted', {
+      query: s.school_name.toLowerCase(),
+      via: 'suggestion',
+      urn: s.urn,
+      has_postcode: false,
+      filters_active: '',
+      filters_count: 0,
+    });
+    router.push(schoolUrl(s.urn, s.school_name));
+  };
+
+  const handleOmniKeyDown = (e: React.KeyboardEvent<HTMLInputElement>) => {
+    if (!open) return;
+    if (e.key === "ArrowDown") {
+      e.preventDefault();
+      setActiveIndex(activeIndex + 1 >= suggestions.length ? 0 : activeIndex + 1);
+    } else if (e.key === "ArrowUp") {
+      e.preventDefault();
+      setActiveIndex(activeIndex <= 0 ? suggestions.length - 1 : activeIndex - 1);
+    } else if (e.key === "Escape") {
+      close();
+    } else if (e.key === "Enter" && activeIndex >= 0) {
+      // Only when an option is active. With none, the event falls through to
+      // the form's submit handler and searches the typed text, as it does now.
+      e.preventDefault();
+      pickSuggestion(suggestions[activeIndex]);
+    }
+  };
+
   const currentLA = searchParams.get("local_authority") || "";
   const currentType = searchParams.get("school_type") || "";
   const currentPhase = searchParams.get("phase") || "";
@@ -124,6 +183,9 @@ export function FilterBar({
 
   const handleSearchSubmit = (e: React.FormEvent) => {
     e.preventDefault();
+    // The search has been made; the suggestions that led to it are spent.
+    setHasTyped(false);
+    close();
     if (!omniValue.trim()) {
       updateURL({ search: "", postcode: "", radius: "" });
       return;
@@ -226,9 +288,20 @@ export function FilterBar({
             ref={inputRef}
             type="search"
             value={omniValue}
-            onChange={(e) => setOmniValue(e.target.value)}
+            onChange={(e) => { setOmniValue(e.target.value); setHasTyped(true); }}
+            onKeyDown={handleOmniKeyDown}
+            onBlur={close}
             placeholder="School name or postcode"
             className={styles.omniInput}
+            {...(autosuggest ? {
+              role: "combobox",
+              "aria-expanded": open,
+              "aria-controls": suggestId,
+              "aria-autocomplete": "list" as const,
+              "aria-activedescendant":
+                activeIndex >= 0 ? suggestOptionId(suggestId, activeIndex) : undefined,
+              autoComplete: "off",
+            } : {})}
           />
           <button
             type="submit"
@@ -237,6 +310,15 @@ export function FilterBar({
           >
             {isPending ? <div className={styles.spinner}></div> : isHero ? "Search schools" : "Search"}
           </button>
+          {autosuggest && open && (
+            <SuggestList
+              id={suggestId}
+              suggestions={suggestions}
+              activeIndex={activeIndex}
+              onPick={pickSuggestion}
+              onHover={setActiveIndex}
+            />
+          )}
         </div>
         {isHero && (
           <>

nextjs-app/components/Footer.module.css

@@ -22,7 +22,8 @@
 
 .content {
   display: grid;
-  grid-template-columns: 1.6fr 1fr 1fr;
+  /* Brand column plus three link columns: Product, Resources, About. */
+  grid-template-columns: 1.6fr 1fr 1fr 1fr;
   gap: 2rem;
   margin-bottom: 3rem;
 }
@@ -193,6 +194,14 @@
   color: var(--on-sunken);
 }
 
+/* Four columns crush between the tablet range and the 768px collapse, so
+   pair them up first rather than jumping straight to a single column. */
+@media (max-width: 960px) {
+  .content {
+    grid-template-columns: 1fr 1fr;
+  }
+}
+
 @media (max-width: 768px) {
   .container {
     padding: 2rem 1rem 1.5rem;

nextjs-app/components/Footer.tsx

@@ -10,7 +10,17 @@
 import { LogoMark } from './Logo';
 import styles from './Footer.module.css';
 
-export function Footer() {
+/**
+ * Both default to false so a caller that forgets a prop hides the link rather
+ * than pointing it at a page that 404s. Same reasoning as backend/flags.py:
+ * "Every flag defaults to False."
+ */
+interface FooterProps {
+  aboutEnabled?: boolean;
+  blogEnabled?: boolean;
+}
+
+export function Footer({ aboutEnabled = false, blogEnabled = false }: FooterProps = {}) {
   const currentYear = new Date().getFullYear();
 
   return (
@@ -93,6 +103,27 @@ export function Footer() {
               </li>
             </ul>
           </div>
+
+          {/* Dropped entirely when both flags are dark, rather than left as an
+              empty heading: shipping dark means the footer renders as it did
+              before the feature existed. */}
+          {(aboutEnabled || blogEnabled) && (
+            <div className={styles.section}>
+              <h4 className={styles.sectionTitle}>About</h4>
+              <ul className={styles.links}>
+                {/* The only route to a named human. Deliberately not in the nav:
+                    the mobile bottom bar already carries four items, and both of
+                    these are lower intent than any of them. Post bylines link
+                    here too, which is where a reader actually asks the question. */}
+                {aboutEnabled && (
+                  <li><a href="/about" className={styles.link}>Who&apos;s behind this</a></li>
+                )}
+                {blogEnabled && (
+                  <li><a href="/blog" className={styles.link}>Blog</a></li>
+                )}
+              </ul>
+            </div>
+          )}
         </div>
 
         <div className={styles.bottom}>

nextjs-app/components/HomeView.module.css

@@ -90,7 +90,18 @@
   isolation: isolate;
   background: var(--hero-ground);
   border-radius: var(--radius-xl);
-  overflow: hidden;
+  /*
+   * Deliberately NOT overflow: hidden.
+   *
+   * It used to be, to clip the artwork and the scrim to the rounded corners —
+   * and it also clipped the search box's suggestion dropdown, which is 320px
+   * tall against 145px of panel below the input. Roughly half the list was cut
+   * off with no indication anything was missing.
+   *
+   * The two things that actually needed clipping round themselves instead, so
+   * the panel can let a dropdown out. Anything absolutely positioned inside
+   * this panel and taller than the space below it depends on this.
+   */
 }
 
 .heroContent {
@@ -107,6 +118,10 @@
   position: absolute;
   inset: 0;
   z-index: 0;
+  /* Rounds itself, because the panel no longer clips it. inset: 0 makes this
+     exactly the panel's own corners. */
+  border-radius: inherit;
+  overflow: hidden;
 }
 
 .heroArt picture,
@@ -143,6 +158,9 @@
   inset: 0;
   z-index: 1;
   pointer-events: none;
+  /* Same reason as .heroArt: the panel stopped clipping, so the scrim keeps
+     its own corners rather than squaring off over the panel's. */
+  border-radius: inherit;
   background: linear-gradient(
     to right,
     var(--hero-ground) 0%,
@@ -331,6 +349,10 @@
     position: static;
     order: -1;
     height: 13rem;
+    /* Top corners only. Here the artwork is a band flush with the top of the
+       panel, not a layer covering it — inheriting all four would leave it
+       floating with rounded bottom corners against the copy below. */
+    border-radius: var(--radius-xl) var(--radius-xl) 0 0;
   }
   /* The band crop puts the schoolhouse at 73% across — reported by
      scripts/build-hero-images.js, which derives it from the crop box rather