backend/app.py

@@ -1232,6 +1232,13 @@ async def get_place(request: Request, kind: str, slug: str,
         "place": {"kind": place.kind, "slug": place.slug, "name": place.name,
                   "count": len(place.urns),
                   "parent_authority": place.parent_authority,
+                  # Every authority the place meaningfully sits in. SW19 is
+                  # mostly Merton but partly Wandsworth; naming one asserts
+                  # something false.
+                  "authorities": [
+                      {"name": name, "slug": _slugify(name), "count": n}
+                      for name, n in place.authorities
+                  ],
                   # Only phases that clear the threshold, so the page links
                   # variants that exist rather than 404s.
                   "phases": [ph for ph in ("primary", "secondary")

backend/places.py

@@ -30,6 +30,12 @@ class Place:
     name: str
     urns: tuple[int, ...]
     parent_authority: str | None    # authority NAME, for the 301 target
+    # Every authority the place meaningfully sits in, largest first. A quarter
+    # of outcodes and a third of towns straddle a boundary — SW19 is mostly
+    # Merton but partly Wandsworth — so naming only one asserts something
+    # false. parent_authority stays single because a redirect needs one
+    # target; this is what the page shows.
+    authorities: tuple[tuple[str, int], ...] = ()
     # URNs per phase, so the per-phase threshold can be applied without
     # re-querying. A place with 30 primaries and 2 secondaries publishes a
     # primary variant and no secondary one.
@@ -71,39 +77,100 @@ def _phase_urns(group, publishable: set[int]) -> dict[str, tuple[int, ...]]:
     return out
 
 
-def _parent_authority(group) -> str | None:
+# A place is described by an authority when it holds at least a tenth of the
-    """The most common authority in a group — the useful 301 target.
+# schools, and at least two. GIAS carries occasional postcode errors — EN6
+# lists two Shropshire schools among fourteen in Hertfordshire — and a bare
+# "any authority present" rule would print those as though they were real.
+# There is deliberately no cap on how many are named. An earlier cut stopped
+# at three, which silently dropped the fourth in exactly the case where the
+# information matters most — a genuinely fragmented place. The share rule is
+# the only limit, and it already bounds the list at ten.
+_AUTHORITY_MIN_SHARE = 0.10
+_AUTHORITY_MIN_SCHOOLS = 2
+
+
+def _authorities(group) -> tuple[tuple[str, int], ...]:
+    """Authorities this place meaningfully sits in, largest first."""
+    from backend.app import EXCLUDED_FILTER_VALUES
 
-    A town spanning several authorities has no single parent, so the mode is
-    the honest answer rather than an arbitrary first row.
-    """
     if "local_authority" not in group.columns:
-        return None
+        return ()
-    top = group["local_authority"].dropna()
+    counts = group["local_authority"].dropna().value_counts()
-    return str(top.mode().iloc[0]) if not top.empty else None
+    total = int(counts.sum())
+    if not total:
+        return ()
+
+    kept = [
+        (str(name), int(n)) for name, n in counts.items()
+        if str(name) not in EXCLUDED_FILTER_VALUES
+        and n >= _AUTHORITY_MIN_SCHOOLS
+        and n / total >= _AUTHORITY_MIN_SHARE
+    ]
+    # A place too small or too fragmented for the share rule still names its
+    # largest authority, or the page would say nothing about where it is.
+    if not kept:
+        for name, n in counts.items():
+            if str(name) not in EXCLUDED_FILTER_VALUES:
+                return ((str(name), int(n)),)
+        return ()
+    return tuple(kept)
+
+
+def _parent_authority(authorities: tuple[tuple[str, int], ...]) -> str | None:
+    """The 301 target: the largest authority a place sits in.
+
+    Derived from `authorities` rather than computed separately. The first cut
+    used `mode()` here while `authorities` used `value_counts()`, and on an
+    exact tie pandas does not guarantee the two pick the same name — so the
+    redirect could have pointed somewhere other than the authority the page
+    named first. One computation, one answer.
+
+    Deriving it also inherits the sentinel filter, so a place can no longer
+    redirect to /schools/authority/does-not-apply.
+    """
+    return authorities[0][0] if authorities else None
 
 
 def _group(df, column: str, kind: str, publishable: set[int]) -> dict[str, Place]:
-    """One Place per distinct value of `column` that clears the threshold."""
+    """One Place per distinct SLUG in `column` that clears the threshold.
+
+    Grouped by slug, not by raw value, because GIAS spells the same place
+    several ways and they all resolve to one URL. Five town slugs come from
+    more than one spelling: "London" (1,819 schools) and "LONDON" (12) both
+    slugify to `london`; Weston-super-Mare is split 14/19 across two
+    spellings; Newcastle-under-Lyme across three.
+
+    Grouping by raw value meant the later group simply overwrote the earlier
+    one in this dict — so /schools/london could have shown twelve schools
+    instead of 1,819, silently and depending on row order.
+
+    The display name is the most common spelling, which is the one a reader
+    expects to see.
+    """
     from backend.app import _slugify
 
     if column not in df.columns:
         return {}
 
+    working = df.assign(_slug=df[column].map(
+        lambda v: _slugify(str(v).strip()) if isinstance(v, str) and v.strip() else None))
+    working = working[working["_slug"].notna() & (working["_slug"] != "")]
+
     out: dict[str, Place] = {}
-    for name, group in df.groupby(column, dropna=True):
+    for slug, group in working.groupby("_slug"):
-        name = str(name).strip()
+        slug = str(slug)
-        if not name:
-            continue
         urns = tuple(sorted({int(u) for u in group["urn"]} & publishable))
         if len(urns) < MIN_SCHOOLS:
             continue
-        slug = _slugify(name)
+        spellings = group[column].dropna().value_counts()
-        if not slug:
+        if spellings.empty:
             continue
+        name = str(spellings.index[0]).strip()
+        authorities = () if kind == "authority" else _authorities(group)
         place = Place(
             kind=kind, slug=slug, name=name, urns=urns,
-            parent_authority=_parent_authority(group) if kind == "town" else None,
+            parent_authority=_parent_authority(authorities),
+            authorities=authorities,
             phase_urns=_phase_urns(group, publishable),
         )
         out[place.key] = place
@@ -136,8 +203,10 @@ def _outcode_places(df, publishable: set[int]) -> dict[str, Place]:
         urns = tuple(sorted({int(u) for u in group["urn"]} & publishable))
         if len(urns) < MIN_SCHOOLS:
             continue
+        authorities = _authorities(group)
         place = Place(kind="outcode", slug=str(oc).lower(), name=str(oc),
-                      urns=urns, parent_authority=_parent_authority(group),
+                      urns=urns, parent_authority=_parent_authority(authorities),
+                      authorities=authorities,
                       phase_urns=_phase_urns(group, publishable))
         out[place.key] = place
     return out
@@ -179,8 +248,10 @@ def _locality_places(df, publishable: set[int],
                 "threshold of %d - not published",
                 slug, ", ".join(outcodes), len(urns), MIN_SCHOOLS)
             continue
+        authorities = _authorities(group)
         place = Place(kind="locality", slug=slug, name=name, urns=urns,
-                      parent_authority=_parent_authority(group),
+                      parent_authority=_parent_authority(authorities),
+                      authorities=authorities,
                       phase_urns=_phase_urns(group, publishable))
         out[place.key] = place
     return out

backend/tests/test_places.py

@@ -229,3 +229,117 @@ def test_no_curated_locality_names_a_london_borough():
         f"these are boroughs, not districts: {sorted(named)} - they already "
         "have an authority page covering every school"
     )
+
+
+def test_a_place_names_every_authority_it_straddles():
+    """SW19 is mostly Merton but partly Wandsworth.
+
+    A quarter of viable outcodes and a third of viable towns cross an
+    authority boundary, so naming only the largest asserts something false.
+    """
+    rows = (_town(26, "London", "Merton", start=300000)
+            + _town(7, "London", "Wandsworth", start=400000))
+    for r in rows:
+        r["postcode"] = "SW19 1AA"
+    reg = build_place_registry(_df(rows))
+
+    names = [n for n, _ in reg["outcode:sw19"].authorities]
+    assert names == ["Merton", "Wandsworth"]          # largest first
+    assert dict(reg["outcode:sw19"].authorities)["Wandsworth"] == 7
+
+
+def test_the_redirect_target_stays_a_single_authority():
+    # parent_authority and authorities do different jobs: a 301 needs one
+    # target, the page needs the truth.
+    rows = (_town(26, "London", "Merton", start=300000)
+            + _town(7, "London", "Wandsworth", start=400000))
+    for r in rows:
+        r["postcode"] = "SW19 1AA"
+    reg = build_place_registry(_df(rows))
+    assert reg["outcode:sw19"].parent_authority == "Merton"
+
+
+def test_a_stray_authority_below_the_share_threshold_is_not_named():
+    # GIAS carries postcode errors — EN6 lists two Shropshire schools among
+    # fourteen in Hertfordshire. Printing those as though real would be worse
+    # than omitting them.
+    rows = (_town(30, "Barnet", "Hertfordshire", start=300000)
+            + _town(1, "Barnet", "Shropshire", start=400000))
+    for r in rows:
+        r["postcode"] = "EN6 1AA"
+    reg = build_place_registry(_df(rows))
+    assert [n for n, _ in reg["outcode:en6"].authorities] == ["Hertfordshire"]
+
+
+def test_a_sentinel_authority_is_never_named():
+    rows = (_town(20, "London", "Merton", start=300000)
+            + _town(6, "London", "Does not apply", start=400000))
+    for r in rows:
+        r["postcode"] = "SW19 1AA"
+    reg = build_place_registry(_df(rows))
+    assert [n for n, _ in reg["outcode:sw19"].authorities] == ["Merton"]
+
+
+def test_a_place_always_names_at_least_one_authority():
+    # Even when every authority is below the share threshold, the page has to
+    # say where the place is.
+    rows = []
+    for i, la in enumerate(["A", "B", "C", "D", "E", "F", "G"]):
+        rows += _town(1, "Fragmented", la, start=300000 + i * 100)
+    reg = build_place_registry(_df(rows))
+    place = reg.get("town:fragmented")
+    assert place is not None
+    assert len(place.authorities) == 1
+
+
+def test_every_qualifying_authority_is_named_with_no_cap():
+    """An earlier cut stopped at three, dropping the fourth silently.
+
+    That truncation bit exactly where the information matters most — a
+    genuinely fragmented place — and nothing recorded it.
+    """
+    rows = []
+    for i, la in enumerate(["Hackney", "Lambeth", "Westminster", "Lewisham"]):
+        rows += _town(3, "Fourway", la, start=300000 + i * 100)
+    reg = build_place_registry(_df(rows))
+    assert len(reg["town:fourway"].authorities) == 4
+
+
+def test_the_redirect_target_is_the_authority_named_first():
+    """They were computed separately — mode() against value_counts() — and on
+    an exact tie pandas does not guarantee the two agree."""
+    rows = (_town(26, "London", "Merton", start=300000)
+            + _town(7, "London", "Wandsworth", start=400000))
+    for r in rows:
+        r["postcode"] = "SW19 1AA"
+    place = build_place_registry(_df(rows))["outcode:sw19"]
+    assert place.parent_authority == place.authorities[0][0]
+
+
+def test_a_place_never_redirects_to_a_sentinel_authority():
+    # Deriving the parent from `authorities` inherits its sentinel filter.
+    rows = (_town(6, "Someplace", "Does not apply", start=300000)
+            + _town(5, "Someplace", "Essex", start=400000))
+    reg = build_place_registry(_df(rows))
+    assert reg["town:someplace"].parent_authority == "Essex"
+
+
+def test_spellings_of_one_place_are_merged_not_overwritten():
+    """GIAS spells the same place several ways, and they share a URL.
+
+    "London" (1,819 schools) and "LONDON" (12) both slugify to `london`.
+    Grouping by raw value let the later group overwrite the earlier one, so
+    the page could have shown twelve schools instead of 1,819 — silently, and
+    depending on row order.
+    """
+    rows = (_town(6, "Weston-super-Mare", "North Somerset", start=300000)
+            + _town(5, "Weston-Super-Mare", "North Somerset", start=400000))
+    reg = build_place_registry(_df(rows))
+    assert len(reg["town:weston-super-mare"].urns) == 11
+
+
+def test_the_merged_place_takes_its_most_common_spelling():
+    rows = (_town(9, "Newcastle-under-Lyme", "Staffordshire", start=300000)
+            + _town(5, "NEWCASTLE-UNDER-LYME", "Staffordshire", start=400000))
+    reg = build_place_registry(_df(rows))
+    assert reg["town:newcastle-under-lyme"].name == "Newcastle-under-Lyme"

e2e/tests/journeys.spec.ts

@@ -1885,3 +1885,29 @@ test('no page title repeats the brand', async ({ page }) => {
     expect(brands, `${path} repeats the brand: ${title}`).toBeLessThanOrEqual(1);
   }
 });
+
+test('a place straddling a boundary names every authority it sits in', async ({ page }) => {
+  // A quarter of outcodes and a third of towns cross an authority boundary —
+  // SW19 is mostly Merton but partly Wandsworth. Naming only the largest
+  // asserts something false about the place.
+  const { places } = await (await page.request.get('/api/places')).json();
+  const outcode = places.find((p: { kind: string }) => p.kind === 'outcode');
+  expect(outcode).toBeTruthy();
+
+  // Find any place the registry reports as straddling.
+  let straddling: { kind: string; slug: string } | null = null;
+  for (const p of places.filter((p: { kind: string }) => p.kind === 'outcode').slice(0, 40)) {
+    const d = await (await page.request.get(`/api/places/outcode/${p.slug}`)).json();
+    if ((d.place.authorities ?? []).length > 1) { straddling = p; break; }
+  }
+  test.skip(!straddling, 'no straddling outcode found in the sample');
+
+  const detail = await (await page.request.get(
+    `/api/places/outcode/${straddling!.slug}`)).json();
+  await page.goto(`/schools/near/${straddling!.slug}`);
+
+  for (const a of detail.place.authorities) {
+    await expect(page.locator(`a[href="/schools/authority/${a.slug}"]`).first())
+      .toBeVisible();
+  }
+});

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

@@ -199,3 +199,47 @@ describe('PlaceView table alignment', () => {
     expect(container.querySelectorAll('th')[0].className).toBe('');
   });
 });
+
+describe('PlaceView authorities', () => {
+  const straddling: PlaceDetail = {
+    place: { kind: 'outcode', slug: 'sw19', name: 'SW19', count: 33,
+             parent_authority: 'Merton', phases: ['primary'],
+             authorities: [
+               { name: 'Merton', slug: 'merton', count: 26 },
+               { name: 'Wandsworth', slug: 'wandsworth', count: 7 },
+             ] },
+    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 every authority the place straddles, not just the largest', () => {
+    // SW19 is mostly Merton but partly Wandsworth. Naming one asserts
+    // something false about a quarter of outcodes.
+    render(<PlaceView detail={straddling} englandAverage={61} neighbours={[]} />);
+    expect(screen.getByRole('link', { name: 'Merton' }))
+      .toHaveAttribute('href', '/schools/authority/merton');
+    expect(screen.getByRole('link', { name: 'Wandsworth' }))
+      .toHaveAttribute('href', '/schools/authority/wandsworth');
+  });
+
+  it('joins them readably rather than as a bare list', () => {
+    // Asserted on the summary line's whole text: a loose /and/ matcher also
+    // hits "Wandsworth".
+    const { container } = render(<PlaceView detail={straddling}
+      englandAverage={61} neighbours={[]} />);
+    const summary = container.querySelector('header p');
+    expect(summary?.textContent).toContain('Merton and Wandsworth');
+  });
+
+  it('falls back to the single parent when the field is absent', () => {
+    // A cached API response predating the authorities field must not blank
+    // the line entirely.
+    const legacy = { ...straddling,
+      place: { ...straddling.place, authorities: undefined } };
+    render(<PlaceView detail={legacy} englandAverage={61} neighbours={[]} />);
+    expect(screen.getByRole('link', { name: 'Merton' })).toBeInTheDocument();
+  });
+});

nextjs-app/components/places/PlaceView.tsx

@@ -106,6 +106,13 @@ function SchoolTable({ schools, phase }: { schools: School[]; phase: PhaseKey })
 
 export function PlaceView({ detail, phase, englandAverage, neighbours }: Props) {
   const { place, schools, averages } = detail;
+  // Fall back to the single parent when the API predates the authorities
+  // field, so a stale cache never blanks the line entirely.
+  const authorities = place.authorities?.length
+    ? place.authorities
+    : place.parent_authority
+      ? [{ name: place.parent_authority, slug: authoritySlug(place.parent_authority), count: 0 }]
+      : [];
   const local = averages[METRICS[phase ?? 'primary'].key];
   const phaseWord = phase === 'secondary' ? 'Secondary schools'
     : phase === 'primary' ? 'Primary schools' : 'Schools';
@@ -163,13 +170,20 @@ export function PlaceView({ detail, phase, englandAverage, neighbours }: Props)
         <h1>{phaseWord} in {place.name}</h1>
         <p className={styles.summary}>
           {place.count} schools
-          {place.parent_authority && (
+          {authorities.length > 0 && (
             <>
               {' · '}
-              <Link href={`/schools/authority/${authoritySlug(place.parent_authority)}`}
+              {/* Every authority, not just the largest. A quarter of outcodes
-                className={styles.inlineLink}>
+                  and a third of towns cross a boundary: SW19 is mostly Merton
-                {place.parent_authority}
+                  but partly Wandsworth, and naming one asserts otherwise. */}
-              </Link>
+              {authorities.map((a, i) => (
+                <span key={a.slug}>
+                  {i > 0 && (i === authorities.length - 1 ? ' and ' : ', ')}
+                  <Link href={`/schools/authority/${a.slug}`} className={styles.inlineLink}>
+                    {a.name}
+                  </Link>
+                </span>
+              ))}
             </>
           )}
         </p>

nextjs-app/lib/places.ts

@@ -18,8 +18,19 @@ export interface PlaceSummary {
   phases?: string[];
 }
 
+export interface PlaceAuthority {
+  name: string;
+  slug: string;
+  count: number;
+}
+
 export interface PlaceDetail {
-  place: PlaceSummary & { parent_authority: string | null };
+  place: PlaceSummary & {
+    parent_authority: string | null;
+    /** Every authority the place meaningfully sits in, largest first. SW19 is
+     *  mostly Merton but partly Wandsworth. */
+    authorities?: PlaceAuthority[];
+  };
   schools: School[];
   averages: {
     rwm_expected_pct: number | null;