diff --git a/applications/modelling_e2e/handler.py b/applications/modelling_e2e/handler.py
index 0f2c4fe6..6e3dece5 100644
--- a/applications/modelling_e2e/handler.py
+++ b/applications/modelling_e2e/handler.py
@@ -453,22 +453,29 @@ def handler(body: dict[str, Any], context: Any) -> Optional[dict[str, Any]]:
                 f"{[s['certificate_number'] for s in skipped_certs]}"
             )
 
-        if failures:
-            failed_ids = [f["property_id"] for f in failures]
-            # Persisted verbatim into the subtask's outputs.error (via
-            # SubTask.fail), so include each property's error type + message —
-            # not just the IDs — to make failed runs diagnosable without
-            # cross-referencing CloudWatch.
-            message = (
-                f"failed property_ids: {failed_ids}; "
-                f"details: {json.dumps(failures)}"
-            )
-            if skipped_certs:
-                message += (
-                    f"; skipped_unmappable_cohort_certs: {json.dumps(skipped_certs)}"
+        # A property that errored AND a cohort cert the mapper could not consume
+        # are both surfaced as failures, so the subtask is marked failed and
+        # shows up for debugging. The whole batch has already run by this point —
+        # every property that could be modelled was written to DB above — so
+        # failing here flags the run without discarding the work done so far.
+        if failures or skipped_certs:
+            parts: list[str] = []
+            if failures:
+                failed_ids = [f["property_id"] for f in failures]
+                # Persisted verbatim into the subtask's outputs.error (via
+                # SubTask.fail): include each property's error type + message,
+                # not just the IDs, so failed runs are diagnosable without
+                # cross-referencing CloudWatch.
+                parts.append(
+                    f"failed property_ids: {failed_ids}; "
+                    f"details: {json.dumps(failures)}"
                 )
-            raise RuntimeError(message)
+            if skipped_certs:
+                parts.append(
+                    f"skipped_unmappable_cohort_certs: {json.dumps(skipped_certs)}"
+                )
+            raise RuntimeError("; ".join(parts))
 
-        return {"skipped_unmappable_cohort_certs": skipped_certs} if skipped_certs else None
+        return None
     finally:
         read_session.close()
diff --git a/tests/applications/modelling_e2e/test_handler.py b/tests/applications/modelling_e2e/test_handler.py
index 6929d13f..540d6040 100644
--- a/tests/applications/modelling_e2e/test_handler.py
+++ b/tests/applications/modelling_e2e/test_handler.py
@@ -214,16 +214,19 @@ def test_lodged_epc_path_saves_epc_plan_and_marks_modelled(
     _baseline_orchestrator.return_value.run.assert_called_once_with([PROPERTY_ID])
 
 
-def test_skipped_cohort_certs_are_surfaced_in_the_outputs() -> None:
+def test_skipped_cohort_certs_fail_the_subtask_but_the_plan_is_still_saved() -> None:
     """Cohort certs the mapper can't consume are skipped (so prediction is not
-    aborted) and surfaced — with cert numbers — in the subtask outputs, so the
-    mapper gaps can be reported and closed."""
+    aborted), then surfaced as a failure — the subtask is marked failed (the
+    cert numbers land in outputs.error via the raised RuntimeError) so the
+    mapper gaps get debugged. The batch still ran to completion first, so the
+    property's plan was committed before the handler raised."""
     from repositories.comparable_properties.epc_comparable_properties_repository import (
         SkippedCohortCert,
     )
 
     mock_engine = _engine_mock([PROPERTY_ID], [UPRN], [POSTCODE])
     mock_plan = _plan_mock()
+    mock_uow = MagicMock()
     skipped = [
         SkippedCohortCert(
             certificate_number="8257-7539-1649-0633-4992",
@@ -287,24 +290,21 @@ def test_skipped_cohort_certs_are_surfaced_in_the_outputs() -> None:
         MockUoW = stack.enter_context(
             patch("applications.modelling_e2e.handler.PostgresUnitOfWork")
         )
-        MockUoW.return_value.__enter__.return_value = MagicMock()
+        MockUoW.return_value.__enter__.return_value = mock_uow
         MockUoW.return_value.__exit__.return_value = False
 
-        # Act
-        result = _call_handler(_BODY)
+        # Act — the skipped cert fails the subtask, but only after the batch ran.
+        with pytest.raises(RuntimeError) as excinfo:
+            _call_handler(_BODY)
 
-    # Assert — the handler's return (→ subtask outputs.result) carries the cert
-    # numbers + errors of every skipped cohort cert.
-    assert result == {
-        "skipped_unmappable_cohort_certs": [
-            {
-                "certificate_number": "8257-7539-1649-0633-4992",
-                "error": (
-                    "ValueError: RdSapSchema17_1: missing required field 'window'"
-                ),
-            }
-        ]
-    }
+    # Assert — the cert number + error reach outputs.error (the raised message),
+    # and the property's plan was still committed before the handler raised.
+    message = str(excinfo.value)
+    assert "skipped_unmappable_cohort_certs" in message
+    assert "8257-7539-1649-0633-4992" in message
+    assert "RdSapSchema17_1: missing required field 'window'" in message
+    mock_uow.plan.save.assert_called_once()
+    mock_uow.commit.assert_called_once()
 
 
 # ---------------------------------------------------------------------------