Let's say a data load has occurred but the data was loaded incorrectly. How would you identify and correct the issue?

First, I would verify the load by comparing source and target row counts, checksums, and key metrics. If discrepancies exist, I would examine the ETL logs for errors or warnings. Next, I would isolate the affected rows using a checksum or hash of the source data and query the target to find mismatches. Once identified, I would rollback the affected portion of the load—either by deleting the bad rows or restoring from a pre‑load snapshot. Then I would re‑run the ETL for that subset, ensuring idempotency by using staging tables and upsert logic. Finally, I would implement validation rules (e.g., data type checks, referential integrity) and automated tests to catch similar issues before production loads. Continuous monitoring and alerting on load failures help prevent recurrence.