Persist QA feature evidence

CHANGELOG.md

@@ -7,6 +7,14 @@
 
 # Changelog
 
+## Sprint 111 QA feature evidence persistence (2026-06-25)
+
+- Added feature-level QA evidence to dataset, detection and segmentation QA matching.
+- Persisted matched feature ids, false-positive feature ids and false-negative feature ids inside `quality_checks.findings_json`.
+- Extended QA/QC drilldown with compact matched/false-positive/false-negative feature id lists beside the existing metrics and raw findings JSON.
+- Updated API contracts to document `match_evidence`, `false_positive_evidence` and `false_negative_evidence`.
+- No migration, new table, provider fetching, AI behavior or new product domain was introduced.
+
 ## Sprint 110 Map QA evidence drilldown (2026-06-25)
 
 - Extended the Map workspace QA/QC shortcut with inline evidence after comparing a saved derived selection dataset.

backend/app/api/routes/qa.py

@@ -59,6 +59,9 @@ def compare_candidate_with_reference(
                 "warnings": result_json.get("warnings", []),
                 "unsupported_geometry": result_json.get("unsupported_geometry", False),
                 "unsupported_geometries": result_json.get("unsupported_geometries", []),
+                "match_evidence": result_json.get("match_evidence", []),
+                "false_positive_evidence": result_json.get("false_positive_evidence", []),
+                "false_negative_evidence": result_json.get("false_negative_evidence", []),
             },
             metrics={
                 "precision": result_json.get("precision"),

backend/app/schemas/qa.py

@@ -28,6 +28,9 @@ class QaProviderComparisonResult(BaseModel):
     iou_threshold: float
     unsupported_geometry: bool = False
     unsupported_geometries: list[str] = Field(default_factory=list)
+    match_evidence: list[dict] = Field(default_factory=list)
+    false_positive_evidence: list[dict] = Field(default_factory=list)
+    false_negative_evidence: list[dict] = Field(default_factory=list)
     generated_at: datetime
 
 

backend/app/services/detection_service.py

@@ -287,18 +287,18 @@ class DetectionService:
 
         candidate_geometries = [({"id": str(row.id), "class_name": row.class_name}, to_shape(row.geometry)) for row in detections]
         reference_geometries = [({"id": str(row.id), "feature_class": row.feature_class}, to_shape(row.geometry)) for row in references]
-        matches, false_positives, false_negatives, match_iou_values, warnings, unsupported = QaService._match_io_u_metrics(
+        evidence = QaService._match_io_u_evidence(
             candidate_geometries,
             reference_geometries,
             iou_threshold,
         )
-        mean_iou = None if not match_iou_values else sum(match_iou_values) / len(match_iou_values)
-        precision = matches / (matches + false_positives) if matches + false_positives > 0 else None
-        recall = matches / (matches + false_negatives) if matches + false_negatives > 0 else None
+        mean_iou = None if not evidence.match_iou_values else sum(evidence.match_iou_values) / len(evidence.match_iou_values)
+        precision = evidence.matches / (evidence.matches + evidence.false_positives) if evidence.matches + evidence.false_positives > 0 else None
+        recall = evidence.matches / (evidence.matches + evidence.false_negatives) if evidence.matches + evidence.false_negatives > 0 else None
         f1_score = None
         if precision is not None and recall is not None:
             f1_score = (2 * precision * recall) / (precision + recall) if precision + recall > 0 else 0.0
-        status = "unsupported" if unsupported else "ok"
+        status = "unsupported" if evidence.unsupported else "ok"
         quality_check = QualityService.persist_quality_check(
             db=db,
             project_id=run.project_id,
@@ -316,19 +316,22 @@ class DetectionService:
                 "min_confidence": min_confidence,
             },
             findings={
-                "matches": matches,
-                "false_positives": false_positives,
-                "false_negatives": false_negatives,
-                "warnings": warnings,
-                "unsupported_geometry": unsupported,
+                "matches": evidence.matches,
+                "false_positives": evidence.false_positives,
+                "false_negatives": evidence.false_negatives,
+                "warnings": evidence.warnings,
+                "unsupported_geometry": evidence.unsupported,
+                "match_evidence": evidence.match_evidence,
+                "false_positive_evidence": evidence.false_positive_evidence,
+                "false_negative_evidence": evidence.false_negative_evidence,
             },
             metrics={
                 "precision": precision,
                 "recall": recall,
                 "f1": f1_score,
                 "mean_iou": mean_iou,
-                "false_positive_count": false_positives,
-                "false_negative_count": false_negatives,
+                "false_positive_count": evidence.false_positives,
+                "false_negative_count": evidence.false_negatives,
             },
         )
         return {
@@ -338,15 +341,18 @@ class DetectionService:
             "reference_dataset_id": str(reference_dataset_id),
             "candidate_feature_count": len(candidate_geometries),
             "reference_feature_count": len(reference_geometries),
-            "matches": matches,
-            "false_positives": false_positives,
-            "false_negatives": false_negatives,
+            "matches": evidence.matches,
+            "false_positives": evidence.false_positives,
+            "false_negatives": evidence.false_negatives,
             "precision": precision,
             "recall": recall,
             "f1_score": f1_score,
             "mean_iou": mean_iou,
             "iou_threshold": iou_threshold,
-            "warnings": warnings,
+            "warnings": evidence.warnings,
+            "match_evidence": evidence.match_evidence,
+            "false_positive_evidence": evidence.false_positive_evidence,
+            "false_negative_evidence": evidence.false_negative_evidence,
         }
 
     @staticmethod

backend/app/services/qa_service.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from typing import Any
 from uuid import UUID
@@ -16,6 +17,19 @@ from app.schemas.qa import QaProviderComparisonResult
 from app.services.vector_operations_service import VectorOperationsService
 
 
+@dataclass
+class QaMatchEvidence:
+    matches: int = 0
+    false_positives: int = 0
+    false_negatives: int = 0
+    match_iou_values: list[float] = field(default_factory=list)
+    warnings: list[str] = field(default_factory=list)
+    unsupported: bool = False
+    match_evidence: list[dict[str, Any]] = field(default_factory=list)
+    false_positive_evidence: list[dict[str, Any]] = field(default_factory=list)
+    false_negative_evidence: list[dict[str, Any]] = field(default_factory=list)
+
+
 def _extract_crs_warnings(source_dataset: Dataset, reference_dataset: Dataset) -> list[str]:
     warnings: list[str] = []
     for dataset, label in ((source_dataset, "candidate"), (reference_dataset, "reference")):
@@ -92,17 +106,34 @@ class QaService:
         return area_geometry
 
     @staticmethod
-    def _match_io_u_metrics(
+    def _feature_identifier(feature: dict[str, Any], fallback_prefix: str, index: int) -> str:
+        feature_id = feature.get("id")
+        if feature_id is not None:
+            return str(feature_id)
+        properties = feature.get("properties")
+        if isinstance(properties, dict):
+            for key in ("vector_feature_id", "source_feature_id", "detection_id", "segmentation_id", "id", "name"):
+                value = properties.get(key)
+                if value is not None:
+                    return str(value)
+        for key in ("vector_feature_id", "source_feature_id", "detection_id", "segmentation_id", "class_name", "feature_class"):
+            value = feature.get(key)
+            if value is not None:
+                return str(value)
+        return f"{fallback_prefix}-{index + 1}"
+
+    @staticmethod
+    def _match_io_u_evidence(
         source_geometries: list[tuple[dict[str, Any], BaseGeometry]],
         reference_geometries: list[tuple[dict[str, Any], BaseGeometry]],
         iou_threshold: float,
-    ) -> tuple[int, int, int, list[float], list[str], bool]:
+    ) -> QaMatchEvidence:
         source_supported = [
-            (feature, geom) for feature, geom in source_geometries if geom.geom_type in QaService.SUPPORTED_GEOMETRY_TYPES
+            (index, feature, geom) for index, (feature, geom) in enumerate(source_geometries) if geom.geom_type in QaService.SUPPORTED_GEOMETRY_TYPES
         ]
         reference_supported = [
-            (feature, geom)
-            for feature, geom in reference_geometries
+            (index, feature, geom)
+            for index, (feature, geom) in enumerate(reference_geometries)
             if geom.geom_type in QaService.SUPPORTED_GEOMETRY_TYPES
         ]
 
@@ -114,29 +145,35 @@ class QaService:
             }
         )
         if not source_supported or not reference_supported:
-            return (
-                0,
-                len(source_supported),
-                len(reference_supported),
-                [],
-                [f"Unsupported geometry types: {unsupported}"] if unsupported else [],
-                True,
+            return QaMatchEvidence(
+                false_positives=len(source_supported),
+                false_negatives=len(reference_supported),
+                warnings=[f"Unsupported geometry types: {unsupported}"] if unsupported else [],
+                unsupported=True,
+                false_positive_evidence=[
+                    {"candidate_feature_id": QaService._feature_identifier(feature, "candidate", source_index)}
+                    for source_index, feature, _ in source_supported
+                ],
+                false_negative_evidence=[
+                    {"reference_feature_id": QaService._feature_identifier(feature, "reference", reference_index)}
+                    for reference_index, feature, _ in reference_supported
+                ],
             )
 
         unmatched_reference_indices = set(range(len(reference_supported)))
-        matches = 0
-        match_iou_values: list[float] = []
-        false_positives = 0
+        evidence = QaMatchEvidence(warnings=[f"Unsupported geometry types: {unsupported}"] if unsupported else [], unsupported=bool(unsupported))
 
-        for _, source_geom in source_supported:
+        for source_index, source_feature, source_geom in source_supported:
+            source_feature_id = QaService._feature_identifier(source_feature, "candidate", source_index)
             if source_geom.area <= 0:
-                false_positives += 1
+                evidence.false_positives += 1
+                evidence.false_positive_evidence.append({"candidate_feature_id": source_feature_id})
                 continue
 
             best_iou = 0.0
             best_index = None
             for reference_index in list(unmatched_reference_indices):
-                _, reference_geom = reference_supported[reference_index]
+                _, _, reference_geom = reference_supported[reference_index]
                 if reference_geom.area <= 0:
                     unmatched_reference_indices.discard(reference_index)
                     continue
@@ -161,16 +198,45 @@ class QaService:
                     best_index = reference_index
 
             if best_index is not None and best_iou >= iou_threshold:
-                matches += 1
-                match_iou_values.append(best_iou)
+                reference_original_index, reference_feature, _ = reference_supported[best_index]
+                evidence.matches += 1
+                evidence.match_iou_values.append(best_iou)
+                evidence.match_evidence.append(
+                    {
+                        "candidate_feature_id": source_feature_id,
+                        "reference_feature_id": QaService._feature_identifier(reference_feature, "reference", reference_original_index),
+                        "iou": best_iou,
+                    }
+                )
                 unmatched_reference_indices.discard(best_index)
             else:
-                false_positives += 1
+                evidence.false_positives += 1
+                evidence.false_positive_evidence.append({"candidate_feature_id": source_feature_id})
 
-        false_negatives = len(unmatched_reference_indices)
-        warnings: list[str] = [f"Unsupported geometry types: {unsupported}"] if unsupported else []
+        evidence.false_negatives = len(unmatched_reference_indices)
+        for reference_index in sorted(unmatched_reference_indices):
+            reference_original_index, reference_feature, _ = reference_supported[reference_index]
+            evidence.false_negative_evidence.append(
+                {"reference_feature_id": QaService._feature_identifier(reference_feature, "reference", reference_original_index)}
+            )
 
-        return matches, false_positives, false_negatives, match_iou_values, warnings, bool(unsupported)
+        return evidence
+
+    @staticmethod
+    def _match_io_u_metrics(
+        source_geometries: list[tuple[dict[str, Any], BaseGeometry]],
+        reference_geometries: list[tuple[dict[str, Any], BaseGeometry]],
+        iou_threshold: float,
+    ) -> tuple[int, int, int, list[float], list[str], bool]:
+        evidence = QaService._match_io_u_evidence(source_geometries, reference_geometries, iou_threshold)
+        return (
+            evidence.matches,
+            evidence.false_positives,
+            evidence.false_negatives,
+            evidence.match_iou_values,
+            evidence.warnings,
+            evidence.unsupported,
+        )
 
     @staticmethod
     def compare_candidate_with_reference(
@@ -206,7 +272,7 @@ class QaService:
             candidate_geometries = QaService._apply_area_filter(candidate_geometries, area_geometry, dataset_id=candidate_dataset.id)
             reference_geometries = QaService._apply_area_filter(reference_geometries, area_geometry, dataset_id=reference_dataset.id)
 
-        matches, false_positives, false_negatives, match_iou_values, warnings, unsupported = QaService._match_io_u_metrics(
+        evidence = QaService._match_io_u_evidence(
             candidate_geometries,
             reference_geometries,
             iou_threshold,
@@ -214,35 +280,38 @@ class QaService:
 
         candidate_feature_count = len(candidate_payload.get("features", [])) if isinstance(candidate_payload, dict) else 0
         reference_feature_count = len(reference_payload.get("features", [])) if isinstance(reference_payload, dict) else 0
-        mean_iou = None if not match_iou_values else sum(match_iou_values) / len(match_iou_values)
+        mean_iou = None if not evidence.match_iou_values else sum(evidence.match_iou_values) / len(evidence.match_iou_values)
 
         precision = None
-        if matches + false_positives > 0:
-            precision = matches / (matches + false_positives)
+        if evidence.matches + evidence.false_positives > 0:
+            precision = evidence.matches / (evidence.matches + evidence.false_positives)
 
         recall = None
-        if matches + false_negatives > 0:
-            recall = matches / (matches + false_negatives)
+        if evidence.matches + evidence.false_negatives > 0:
+            recall = evidence.matches / (evidence.matches + evidence.false_negatives)
 
         f1_score = None
         if precision is not None and recall is not None and precision + recall > 0:
             f1_score = (2 * precision * recall) / (precision + recall)
 
-        status = "unsupported" if unsupported else "ok"
+        status = "unsupported" if evidence.unsupported else "ok"
         return QaProviderComparisonResult(
             status=status,
-            warnings=_extract_crs_warnings(candidate_dataset, reference_dataset) + warnings,
+            warnings=_extract_crs_warnings(candidate_dataset, reference_dataset) + evidence.warnings,
             candidate_feature_count=candidate_feature_count,
             reference_feature_count=reference_feature_count,
-            matches=matches,
-            false_positives=false_positives,
-            false_negatives=false_negatives,
+            matches=evidence.matches,
+            false_positives=evidence.false_positives,
+            false_negatives=evidence.false_negatives,
             precision=precision,
             recall=recall,
             f1_score=f1_score,
             mean_iou=mean_iou,
             iou_threshold=iou_threshold,
-            unsupported_geometry=unsupported,
-            unsupported_geometries=warnings,
+            unsupported_geometry=evidence.unsupported,
+            unsupported_geometries=evidence.warnings,
+            match_evidence=evidence.match_evidence,
+            false_positive_evidence=evidence.false_positive_evidence,
+            false_negative_evidence=evidence.false_negative_evidence,
             generated_at=datetime.now(timezone.utc),
         )

backend/app/services/segmentation_service.py

@@ -251,18 +251,18 @@ class SegmentationService:
 
         candidate_geometries = [({"id": str(row.id), "class_name": row.class_name}, to_shape(row.geometry)) for row in segmentations]
         reference_geometries = [({"id": str(row.id), "feature_class": row.feature_class}, to_shape(row.geometry)) for row in references]
-        matches, false_positives, false_negatives, match_iou_values, warnings, unsupported = QaService._match_io_u_metrics(
+        evidence = QaService._match_io_u_evidence(
             candidate_geometries,
             reference_geometries,
             iou_threshold,
         )
-        mean_iou = None if not match_iou_values else sum(match_iou_values) / len(match_iou_values)
-        precision = matches / (matches + false_positives) if matches + false_positives > 0 else None
-        recall = matches / (matches + false_negatives) if matches + false_negatives > 0 else None
+        mean_iou = None if not evidence.match_iou_values else sum(evidence.match_iou_values) / len(evidence.match_iou_values)
+        precision = evidence.matches / (evidence.matches + evidence.false_positives) if evidence.matches + evidence.false_positives > 0 else None
+        recall = evidence.matches / (evidence.matches + evidence.false_negatives) if evidence.matches + evidence.false_negatives > 0 else None
         f1_score = None
         if precision is not None and recall is not None:
             f1_score = (2 * precision * recall) / (precision + recall) if precision + recall > 0 else 0.0
-        status = "unsupported" if unsupported else "ok"
+        status = "unsupported" if evidence.unsupported else "ok"
         quality_check = QualityService.persist_quality_check(
             db=db,
             project_id=run.project_id,
@@ -280,19 +280,22 @@ class SegmentationService:
                 "min_confidence": min_confidence,
             },
             findings={
-                "matches": matches,
-                "false_positives": false_positives,
-                "false_negatives": false_negatives,
-                "warnings": warnings,
-                "unsupported_geometry": unsupported,
+                "matches": evidence.matches,
+                "false_positives": evidence.false_positives,
+                "false_negatives": evidence.false_negatives,
+                "warnings": evidence.warnings,
+                "unsupported_geometry": evidence.unsupported,
+                "match_evidence": evidence.match_evidence,
+                "false_positive_evidence": evidence.false_positive_evidence,
+                "false_negative_evidence": evidence.false_negative_evidence,
             },
             metrics={
                 "precision": precision,
                 "recall": recall,
                 "f1": f1_score,
                 "mean_iou": mean_iou,
-                "false_positive_count": false_positives,
-                "false_negative_count": false_negatives,
+                "false_positive_count": evidence.false_positives,
+                "false_negative_count": evidence.false_negatives,
             },
         )
         return {
@@ -302,15 +305,18 @@ class SegmentationService:
             "reference_dataset_id": str(reference_dataset_id),
             "candidate_feature_count": len(candidate_geometries),
             "reference_feature_count": len(reference_geometries),
-            "matches": matches,
-            "false_positives": false_positives,
-            "false_negatives": false_negatives,
+            "matches": evidence.matches,
+            "false_positives": evidence.false_positives,
+            "false_negatives": evidence.false_negatives,
             "precision": precision,
             "recall": recall,
             "f1_score": f1_score,
             "mean_iou": mean_iou,
             "iou_threshold": iou_threshold,
-            "warnings": warnings,
+            "warnings": evidence.warnings,
+            "match_evidence": evidence.match_evidence,
+            "false_positive_evidence": evidence.false_positive_evidence,
+            "false_negative_evidence": evidence.false_negative_evidence,
         }
 
     @staticmethod

backend/tests/test_qa_service.py

@@ -22,23 +22,30 @@ class FakeSession:
         return None
 
 
+def _feature(feature_id: str, coordinates: list[list[list[float]]]) -> dict:
+    return {
+        "type": "Feature",
+        "id": feature_id,
+        "properties": {"source_feature_id": feature_id},
+        "geometry": {
+            "type": "Polygon",
+            "coordinates": coordinates,
+        },
+    }
+
+
 def _write_dataset(path: Path, coordinates: list[list[list[float]]]) -> None:
     payload = {
         "type": "FeatureCollection",
-        "features": [
-            {
-                "type": "Feature",
-                "properties": {},
-                "geometry": {
-                    "type": "Polygon",
-                    "coordinates": coordinates,
-                },
-            },
-        ],
+        "features": [_feature("feature-1", coordinates)],
     }
     path.write_text(json.dumps(payload), encoding="utf-8")
 
 
+def _write_features(path: Path, features: list[dict]) -> None:
+    path.write_text(json.dumps({"type": "FeatureCollection", "features": features}), encoding="utf-8")
+
+
 def test_qa_compare_candidate_with_reference_returns_metrics(tmp_path) -> None:
     project_id = uuid4()
     candidate_id = uuid4()
@@ -87,6 +94,62 @@ def test_qa_compare_candidate_with_reference_returns_metrics(tmp_path) -> None:
     assert result.f1_score == 1.0
 
 
+def test_qa_compare_candidate_with_reference_returns_feature_level_evidence(tmp_path) -> None:
+    project_id = uuid4()
+    candidate_id = uuid4()
+    reference_id = uuid4()
+    candidate_path = tmp_path / "candidate.geojson"
+    reference_path = tmp_path / "reference.geojson"
+    matched_candidate = [[[4.0, 51.0], [4.1, 51.0], [4.1, 51.1], [4.0, 51.1], [4.0, 51.0]]]
+    matched_reference = [[[4.0, 51.0], [4.1, 51.0], [4.1, 51.1], [4.0, 51.1], [4.0, 51.0]]]
+    false_positive = [[[4.5, 51.5], [4.6, 51.5], [4.6, 51.6], [4.5, 51.6], [4.5, 51.5]]]
+    false_negative = [[[4.8, 51.8], [4.9, 51.8], [4.9, 51.9], [4.8, 51.9], [4.8, 51.8]]]
+    _write_features(candidate_path, [_feature("candidate-match", matched_candidate), _feature("candidate-extra", false_positive)])
+    _write_features(reference_path, [_feature("reference-match", matched_reference), _feature("reference-missing", false_negative)])
+
+    candidate = Dataset(
+        id=candidate_id,
+        project_id=project_id,
+        name="candidate.geojson",
+        dataset_type="vector",
+        source="test",
+        storage_path=str(candidate_path),
+        crs="EPSG:4326",
+        metadata_json={"crs_assumed": False},
+    )
+    reference = Dataset(
+        id=reference_id,
+        project_id=project_id,
+        name="reference.geojson",
+        dataset_type="vector",
+        source="test",
+        storage_path=str(reference_path),
+        crs="EPSG:4326",
+        metadata_json={"crs_assumed": False},
+    )
+
+    result = QaService.compare_candidate_with_reference(
+        db=FakeSession([candidate, reference]),
+        project_id=project_id,
+        candidate_dataset_id=candidate_id,
+        reference_dataset_id=reference_id,
+        iou_threshold=0.5,
+    )
+
+    assert result.matches == 1
+    assert result.false_positives == 1
+    assert result.false_negatives == 1
+    assert result.match_evidence == [
+        {
+            "candidate_feature_id": "candidate-match",
+            "reference_feature_id": "reference-match",
+            "iou": 1.0,
+        }
+    ]
+    assert result.false_positive_evidence == [{"candidate_feature_id": "candidate-extra"}]
+    assert result.false_negative_evidence == [{"reference_feature_id": "reference-missing"}]
+
+
 def test_dataset_reference_metadata_migration_declares_required_columns() -> None:
     migration_path = Path(__file__).parents[1] / "alembic" / "versions" / "202606120001_add_dataset_reference_metadata.py"
     migration_text = migration_path.read_text(encoding="utf-8")

backend/tests/test_sprint111_qa_feature_evidence.py

@@ -0,0 +1,39 @@
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+def read_text(relative_path: str) -> str:
+    return (ROOT / relative_path).read_text(encoding="utf-8")
+
+
+def test_qa_feature_evidence_contract_is_documented_and_rendered() -> None:
+    schema = read_text("backend/app/schemas/qa.py")
+    quality_panel = read_text("frontend/src/components/quality/QualityResultsPanel.tsx")
+    api_contracts = read_text("docs/API_CONTRACTS.md")
+
+    for field_name in ("match_evidence", "false_positive_evidence", "false_negative_evidence"):
+        assert field_name in schema
+        assert field_name in quality_panel
+        assert field_name in api_contracts
+
+    assert "Feature-level QA/QC evidence" in quality_panel
+    assert "Matched feature ids" in quality_panel
+    assert "False positive feature ids" in quality_panel
+    assert "False negative feature ids" in quality_panel
+    assert "evidenceLabel" in quality_panel
+
+
+def test_qa_services_persist_feature_evidence_without_new_migrations() -> None:
+    qa_route = read_text("backend/app/api/routes/qa.py")
+    detection_service = read_text("backend/app/services/detection_service.py")
+    segmentation_service = read_text("backend/app/services/segmentation_service.py")
+    migrations = "\n".join(path.name for path in (ROOT / "backend" / "alembic" / "versions").glob("*.py"))
+
+    for field_name in ("match_evidence", "false_positive_evidence", "false_negative_evidence"):
+        assert field_name in qa_route
+        assert field_name in detection_service
+        assert field_name in segmentation_service
+
+    assert "quality_check_items" not in migrations

backend/tests/test_sprint7a_persistence_foundation.py

@@ -267,6 +267,15 @@ def test_qa_route_persists_quality_check_domain_record(monkeypatch) -> None:
                     "mean_iou": 1.0,
                     "iou_threshold": 0.5,
                     "warnings": [],
+                    "match_evidence": [
+                        {
+                            "candidate_feature_id": "candidate-1",
+                            "reference_feature_id": "reference-1",
+                            "iou": 1.0,
+                        }
+                    ],
+                    "false_positive_evidence": [{"candidate_feature_id": "candidate-extra"}],
+                    "false_negative_evidence": [{"reference_feature_id": "reference-missing"}],
                 }
             },
         )(),
@@ -287,6 +296,9 @@ def test_qa_route_persists_quality_check_domain_record(monkeypatch) -> None:
     assert persisted_quality_checks[0].job_id == job_id
     assert persisted_quality_checks[0].candidate_dataset_id == candidate_dataset_id
     assert persisted_quality_checks[0].reference_dataset_id == reference_dataset_id
+    assert persisted_quality_checks[0].findings_json["match_evidence"][0]["candidate_feature_id"] == "candidate-1"
+    assert persisted_quality_checks[0].findings_json["false_positive_evidence"][0]["candidate_feature_id"] == "candidate-extra"
+    assert persisted_quality_checks[0].findings_json["false_negative_evidence"][0]["reference_feature_id"] == "reference-missing"
     assert [metric.metric_key for metric in persisted_metrics] == [
         "precision",
         "recall",

docs/API_CONTRACTS.md

@@ -988,6 +988,14 @@ Request:
 
 Response is wrapped in the job envelope. On success, `result_json` includes precision, recall, F1, mean IoU, false positives, false negatives and `quality_check_id`.
 
+Sprint 111 also includes feature-level evidence arrays for map/review handoff:
+
+- `match_evidence`: matched candidate/reference feature ids with IoU.
+- `false_positive_evidence`: unmatched candidate feature ids.
+- `false_negative_evidence`: unmatched reference feature ids.
+
+These arrays are derived from the same persisted/source geometries used for IoU matching. They are not separate QA records yet; they are persisted inside `quality_checks.findings_json`.
+
 Sprint 7A persists the QA/QC result as:
 
 - `jobs`: execution state.
@@ -1016,7 +1024,28 @@ Response:
       "status": "ok",
       "score": 0.5,
       "parameters_json": {},
-      "findings_json": {},
+      "findings_json": {
+        "matches": 1,
+        "false_positives": 1,
+        "false_negatives": 1,
+        "match_evidence": [
+          {
+            "candidate_feature_id": "candidate-feature-id",
+            "reference_feature_id": "reference-feature-id",
+            "iou": 0.83
+          }
+        ],
+        "false_positive_evidence": [
+          {
+            "candidate_feature_id": "candidate-extra-id"
+          }
+        ],
+        "false_negative_evidence": [
+          {
+            "reference_feature_id": "reference-missing-id"
+          }
+        ]
+      },
       "metrics": [
         {
           "metric_key": "precision",

docs/CODEX_EXECUTION_LOG.md

@@ -1,3 +1,37 @@
+## Sprint 111 QA feature evidence persistence (2026-06-25)
+
+Changed:
+- Added feature-level evidence extraction to the shared QA IoU matcher.
+- Dataset QA now returns and persists `match_evidence`, `false_positive_evidence` and `false_negative_evidence`.
+- Detection QA and segmentation QA now use the same evidence-aware matcher and persist the same evidence keys in `quality_checks.findings_json`.
+- Extended the QA/QC drilldown with compact matched, false-positive and false-negative feature id lists before the raw findings JSON.
+- Updated `docs/API_CONTRACTS.md`, `frontend/README.md`, `CHANGELOG.md` and `docs/TODO.md`.
+- Added/extended regression coverage in `backend/tests/test_qa_service.py`, `backend/tests/test_sprint7a_persistence_foundation.py` and `backend/tests/test_sprint111_qa_feature_evidence.py`.
+
+Validation:
+- RED: `python -m pytest backend\tests\test_qa_service.py -q` failed before implementation because `QaProviderComparisonResult` had no `match_evidence`.
+- RED: `python -m pytest backend\tests\test_sprint111_qa_feature_evidence.py -q` failed before docs were updated because `docs/API_CONTRACTS.md` did not document the evidence keys.
+- `python -m pytest backend\tests\test_qa_service.py -q` passed: 3 tests.
+- `python -m pytest backend\tests\test_sprint8c_detection_visualization_qa.py backend\tests\test_sprint9_segmentation_foundation.py -q` passed: 18 tests.
+- `python -m pytest backend\tests\test_sprint111_qa_feature_evidence.py -q` passed: 2 tests.
+- `python -m pytest backend\tests\test_qa_service.py backend\tests\test_sprint7a_persistence_foundation.py -q` passed: 10 tests.
+- `python -m compileall backend/app` passed.
+- `cd backend && python -m pytest -q` passed: 352 tests.
+- `cd frontend && npm run typecheck` passed.
+- `cd frontend && npm run build` passed.
+- `cd backend && python -m alembic heads` passed: `202606120900 (head)`.
+- `cd backend && python -m alembic upgrade head --sql` passed.
+- `bash -n scripts/live_migration_smoke.sh` passed.
+- `bash scripts/run_readiness_check.sh` passed.
+
+Limitations:
+- Feature-level evidence is persisted as ids/IoU metadata in `quality_checks.findings_json`; no `quality_check_items` table or first-class evidence geometry table was introduced.
+- Evidence map overlays can now be built from persisted ids, but overlay generation remains future work.
+- No migration, provider fetching, AI dependency, real model behavior or new product domain was added.
+
+Next recommended pass:
+- Add QA evidence overlay generation by resolving persisted evidence ids back to candidate/reference geometries and rendering false positives/false negatives as MapLibre layers.
+
 ## Sprint 110 Map QA evidence drilldown (2026-06-25)
 
 Changed:

docs/TODO.md

@@ -377,3 +377,4 @@ This file now starts with the current implementation status. Older preparation/b
 - [x] Persist map area selections as reusable derived vector datasets indexed into `vector_features`.
 - [x] Add Map workspace QA/QC shortcut for saved derived selection datasets.
 - [x] Add Map workspace QA/QC evidence drilldown handoff for saved selection comparisons.
+- [x] Persist QA/QC feature-level evidence for matches, false positives and false negatives.

frontend/README.md

@@ -287,6 +287,7 @@ AI Lab run controls explicitly explain when no raster dataset is available, inst
 - Raster controls show the latest generated tile manifest path from persisted `raster.tile` jobs and can hand that path directly to Detection Lab or Segmentation Lab with the selected raster dataset.
 - The QA/QC workspace shows candidate/reference handoff cards and resolves persisted quality-check dataset IDs back to dataset names when the datasets are loaded in the current project context.
 - The QA/QC workspace includes a selected-check evidence drilldown with candidate/reference provenance, false-positive/negative metric evidence, map handoff context and parameters/findings JSON.
+- QA/QC findings now persist feature-level evidence in `findings_json`: matched candidate/reference feature ids with IoU, false-positive candidate feature ids and false-negative reference feature ids. The QA/QC drilldown renders these as compact evidence lists before the raw JSON.
 
 ## Raster dependency visibility
 

frontend/src/components/quality/QualityResultsPanel.tsx

@@ -46,6 +46,20 @@ function metricByKey(check: QualityCheckRead | null, metricKey: string): MetricR
   return check?.metrics.find((metric) => metric.metric_key === metricKey)
 }
 
+function findingEvidenceList(check: QualityCheckRead | null, key: string): Record<string, unknown>[] {
+  const value = check?.findings_json?.[key]
+  return Array.isArray(value) ? value.filter((item): item is Record<string, unknown> => Boolean(item) && typeof item === 'object' && !Array.isArray(item)) : []
+}
+
+function evidenceLabel(item: Record<string, unknown>): string {
+  const candidate = item['candidate_feature_id']
+  const reference = item['reference_feature_id']
+  const iou = item['iou']
+  const pairLabel = [candidate ? `Candidate ${candidate}` : null, reference ? `Reference ${reference}` : null].filter(Boolean).join(' / ')
+  const iouValue = Number(iou)
+  return iou === null || iou === undefined || !Number.isFinite(iouValue) ? pairLabel || JSON.stringify(item) : `${pairLabel || 'Match'} / IoU ${iouValue.toFixed(3)}`
+}
+
 function formatQualityTimestamp(value?: string | null): string {
   return value || 'n/a'
 }
@@ -108,6 +122,9 @@ export function QualityResultsPanel({
   const selectedReferenceName = selectedQualityCheck
     ? datasetNameById.get(selectedQualityCheck.reference_dataset_id) ?? selectedQualityCheck.reference_dataset_id
     : 'n/a'
+  const selectedMatchEvidence = findingEvidenceList(selectedQualityCheck, 'match_evidence')
+  const selectedFalsePositiveEvidence = findingEvidenceList(selectedQualityCheck, 'false_positive_evidence')
+  const selectedFalseNegativeEvidence = findingEvidenceList(selectedQualityCheck, 'false_negative_evidence')
   const qualityStatuses = useMemo(() => Array.from(new Set(qualityChecks.map((check) => check.status))).sort(), [qualityChecks])
   const qualityTypes = useMemo(() => Array.from(new Set(qualityChecks.map((check) => check.check_type))).sort(), [qualityChecks])
   const filteredQualityChecks = useMemo(
@@ -239,6 +256,44 @@ export function QualityResultsPanel({
                   <p>Use the candidate and reference datasets as map layers for spatial review.</p>
                 </div>
               </div>
+              <div className="quality-feature-evidence-grid" aria-label="Feature-level QA/QC evidence">
+                <div>
+                  <span>Matched feature ids</span>
+                  {selectedMatchEvidence.length > 0 ? (
+                    <ul>
+                      {selectedMatchEvidence.slice(0, 6).map((item, index) => (
+                        <li key={`${evidenceLabel(item)}-${index}`}>{evidenceLabel(item)}</li>
+                      ))}
+                    </ul>
+                  ) : (
+                    <p>No matched feature ids persisted for this check.</p>
+                  )}
+                </div>
+                <div>
+                  <span>False positive feature ids</span>
+                  {selectedFalsePositiveEvidence.length > 0 ? (
+                    <ul>
+                      {selectedFalsePositiveEvidence.slice(0, 6).map((item, index) => (
+                        <li key={`${evidenceLabel(item)}-${index}`}>{evidenceLabel(item)}</li>
+                      ))}
+                    </ul>
+                  ) : (
+                    <p>No false-positive feature ids persisted for this check.</p>
+                  )}
+                </div>
+                <div>
+                  <span>False negative feature ids</span>
+                  {selectedFalseNegativeEvidence.length > 0 ? (
+                    <ul>
+                      {selectedFalseNegativeEvidence.slice(0, 6).map((item, index) => (
+                        <li key={`${evidenceLabel(item)}-${index}`}>{evidenceLabel(item)}</li>
+                      ))}
+                    </ul>
+                  ) : (
+                    <p>No false-negative feature ids persisted for this check.</p>
+                  )}
+                </div>
+              </div>
               <div className="quality-provenance-grid">
                 <div>
                   <span>Parameters</span>

frontend/src/styles/app.css

@@ -1618,6 +1618,7 @@ button.entity-card {
 
 .quality-drilldown-grid,
 .quality-evidence-token-grid,
+.quality-feature-evidence-grid,
 .quality-provenance-grid {
   display: grid;
   grid-template-columns: repeat(auto-fit, minmax(12rem, 1fr));
@@ -1641,6 +1642,7 @@ button.entity-card {
 
 .quality-drilldown-grid > div,
 .quality-evidence-token-grid > div,
+.quality-feature-evidence-grid > div,
 .quality-provenance-grid > div {
   min-width: 0;
   border: 1px solid var(--line);
@@ -1663,6 +1665,7 @@ button.entity-card {
 .quality-handoff-grid span,
 .quality-drilldown-grid span,
 .quality-evidence-token-grid span,
+.quality-feature-evidence-grid span,
 .quality-provenance-grid span,
 .quality-score-row span,
 .latest-export-card span,
@@ -1703,7 +1706,8 @@ button.entity-card {
 }
 
 .quality-drilldown-grid p,
-.quality-evidence-token-grid p {
+.quality-evidence-token-grid p,
+.quality-feature-evidence-grid p {
   margin: 0.24rem 0 0;
   color: var(--muted);
   font-size: 0.8rem;
@@ -1711,6 +1715,17 @@ button.entity-card {
   overflow-wrap: anywhere;
 }
 
+.quality-feature-evidence-grid ul {
+  display: grid;
+  gap: 0.28rem;
+  margin: 0.38rem 0 0;
+  padding-left: 1.1rem;
+  color: var(--text);
+  font-size: 0.78rem;
+  line-height: 1.35;
+  overflow-wrap: anywhere;
+}
+
 .quality-provenance-pre {
   max-height: 11rem;
   margin: 0.38rem 0 0;