Preserve provenance
Every output retains the source record, ingestion run and transformation version.
A traceable flow from source records to reporting, with reconciliation and visible exceptions.
Records arrive from disconnected systems with different identifiers, timestamps and status labels. Teams reconcile extracts manually, and discrepancies are difficult to trace.
Map source systems and agree a common record contract. Preserve source identity, event time, amount units, currency, status mappings and the transformation version. Validate the outputs with comparison runs before reporting depends on them.
Every output retains the source record, ingestion run and transformation version.
Missing identities and unknown states enter a review queue. Do not silently discard or reinterpret them.
Use stable identifiers and versions so reloading a source does not duplicate the event. Separate event time from ingestion time.
Proposed acceptance checks. Confirm the actual rules and thresholds with the owner.
Source system and stable source identifier are present.
Amount units, currency and rounding rules remain explicit.
Counts and amounts reconcile by source and reporting period.
The data owner agrees how late events affect prepared periods.
Source register, mapping specification, validation evidence and operating instructions for monitoring, recovery and reruns.
Agree freshness, completeness and reconciliation thresholds with the data owner. Compare investigation effort against the original process. No measured client outcome is asserted.
Discuss similar workA clear brief or a half-formed idea.
Start with what you know.