Data exceptions need an owner and a queue.

Trace the rework

A pipeline may finish while leaving ambiguous records for a person to resolve. If those records reappear on every run, the team pays in processing and repeated investigation. The visible symptom may be a growing exception count, late reporting, or a spreadsheet maintained outside the platform.

An exception record should retain the original input, the reason it is held, the affected business output, and the person able to resolve it. A later decision needs a traceable connection to the corrected result so the same uncertainty does not quietly return.

Choose one recurring exception

The first investigation can follow one recurring exception through its processing attempts and manual handoffs. That evidence helps distinguish a source-quality issue from a missing business rule or recovery design. A Reliability Office can use the pattern to prioritize an improvement and verify whether rework falls after the change.