Data Reconciliation Checklist

When to Use This Checklist

Use this checklist after a data migration to prove that the target holds the same data as the source, completely and accurately. Reconciliation is the formal comparison that turns confidence into evidence. It is essential for financial, regulated, and business-critical data, where an undetected discrepancy can mean lost revenue or a failed audit.

How to Use This Checklist

Define the scope first: which entities, fields, and time windows you will reconcile. Layer the checks from coarse to fine: row counts, then control totals, then row-level checksums. Crucially, account for intended transformations so expected differences are not reported as errors. Pay special attention to boundary records around the cutover timestamp, where in-flight data often goes missing or duplicates. Every discrepancy needs a documented root cause, and data owners must sign off on the result.

What Good Looks Like

Strong reconciliation produces matching row counts, matching control totals, and matching checksums across the defined scope, with every difference explained by a known transformation. Records present in only one system are categorized and resolved rather than ignored. Boundary records around cutover are explicitly checked. The result is documented, signed off by data owners, and ideally automated so it can run repeatedly during a parallel-operation period. Reconciliation becomes audit-ready evidence, not a verbal assurance.

Common Pitfalls

The most common error is reconciling only row counts, which can match while values differ. Teams also flag legitimate transformation differences as discrepancies, eroding trust in the process, or do the opposite and dismiss real errors as expected. Boundary records around the cutover are frequently overlooked, hiding lost or duplicated in-flight data. Finally, reconciliation without owner sign-off carries no weight in an audit, no matter how thorough the checks were.

Related Resources

Ground reconciliation in data-quality-management and use data-lineage to trace discrepancies to their source. Define expected transformations with data-contracts and assign accountability through a data-governance-framework.