diff --git a/backend/services/usage_metrics_service.py b/backend/services/usage_metrics_service.py
index 8b1e09ba6..80297d2d3 100644
--- a/backend/services/usage_metrics_service.py
+++ b/backend/services/usage_metrics_service.py
@@ -297,7 +297,9 @@ def _extract_execution_result(
         if consensus_data is not None and consensus_data.leader_receipt:
             first_receipt = consensus_data.leader_receipt[0]
             if first_receipt is not None:
-                execution_result = getattr(first_receipt, "execution_result", None)
+                execution_result = self._receipt_field(
+                    first_receipt, "execution_result"
+                )
                 if execution_result is not None:
                     # Handle both enum and string values
                     if hasattr(execution_result, "value"):
diff --git a/tests/unit/test_bug_hunt_v0_123_dev.py b/tests/unit/test_bug_hunt_v0_123_dev.py
new file mode 100644
index 000000000..39c536cde
--- /dev/null
+++ b/tests/unit/test_bug_hunt_v0_123_dev.py
@@ -0,0 +1,18 @@
+from types import SimpleNamespace
+
+from backend.services.usage_metrics_service import UsageMetricsService
+
+
+def test_usage_metrics_preserves_error_from_serialized_leader_receipt():
+    """Finalization metrics must not turn a VM error into a successful decision."""
+    consensus_data = SimpleNamespace(
+        leader_receipt=[{"execution_result": "ERROR"}],
+    )
+
+    result = UsageMetricsService()._extract_execution_result(
+        finalization_data={},
+        consensus_data=consensus_data,
+        consensus_history=None,
+    )
+
+    assert result == "error"