Locate the failing boundary
Treat the flow as a sequence of contracts: source extraction, file or payload construction, transport, intake, mapping, processing, and downstream consumption. Confirm the last successful boundary before changing anything.
- Source row count and extraction time
- Filename, path, encoding, and delimiter
- Transport receipt and permissions
- Schema and field mapping
- Processing status and reject evidence
- Downstream record and report checks
Use reconciliation, not visual sampling
A file that looks correct can still have duplicate keys, invalid dates, invisible delimiters, inconsistent quoting, or counts that change after filtering. Reconcile totals and key populations at each boundary.
Check time and scheduling assumptions
Time zone, daylight saving time changes, job overlap, upstream delays, and retry behavior can create intermittent failures that appear to be data issues.
- Declared time zone
- Expected and actual arrival
- Overlap and lock behavior
- Retry and duplicate handling
- Holiday and weekend schedules
Close with operational controls
Once the immediate issue is resolved, add monitoring, ownership, a failure runbook, and reconciliation evidence. A fix without an operating control usually becomes a repeat incident.