diff --git a/pipeline/transform/dbt_project.yml b/pipeline/transform/dbt_project.yml
index 86dce6a..ba23641 100644
--- a/pipeline/transform/dbt_project.yml
+++ b/pipeline/transform/dbt_project.yml
@@ -23,6 +23,11 @@ models:
       +materialized: table
       +schema: marts
 
+tests:
+  school_compare:
+    assert_no_orphaned_facts:
+      +severity: warn
+
 seeds:
   school_compare:
     +schema: seeds
diff --git a/pipeline/transform/models/intermediate/int_ks2_with_lineage.sql b/pipeline/transform/models/intermediate/int_ks2_with_lineage.sql
index e9c4066..68dc193 100644
--- a/pipeline/transform/models/intermediate/int_ks2_with_lineage.sql
+++ b/pipeline/transform/models/intermediate/int_ks2_with_lineage.sql
@@ -19,7 +19,8 @@ with current_ks2 as (
 ),
 
 predecessor_ks2 as (
-    select
+    -- If multiple predecessors have data for the same year, keep the one with most pupils.
+    select distinct on (lin.current_urn, ks2.year)
         lin.current_urn,
         ks2.urn as source_urn,
         ks2.year, ks2.total_pupils, ks2.eligible_pupils,
@@ -40,6 +41,7 @@ predecessor_ks2 as (
         where curr.urn = lin.current_urn
           and curr.year = ks2.year
     )
+    order by lin.current_urn, ks2.year, ks2.total_pupils desc nulls last
 ),
 
 combined as (