RESOURCE · CHECKLIST
Critical-data quality checklist
Reliable does not mean it looks right in a chart. It means the data passes explicit checks that someone owns. Run these against the few datasets your decisions actually depend on.
Free. No email required.
Pick the critical data first
- List the few datasets that directly feed decisions or external reporting. Check those, not everything.
Run six checks on each
- Completeness: are required fields and expected rows all present?
- Uniqueness: are there unintended duplicate business keys?
- Validity: do values conform to the expected type, range and allowed set?
- Referential integrity: does every child row point to a valid parent key?
- Freshness: is the data recent enough for the decision it supports?
- Reconciliation: do independent sources agree at record level, not just on totals?
- Record the owner for each dataset and where failures are alerted.
Why this works, in depth: How do we know whether our data is actually reliable?