The six metrics
Define each one per critical field, not for the database as a whole.
- Completeness: share of records with the field populated
- Accuracy: share matching a trusted reference on sample audit
- Validity: share conforming to the format or allowed value list
- Consistency: share agreeing across systems that hold the same field
- Timeliness: age since last verification against your target
- Uniqueness: duplicate rate under a documented match rule
Sampling makes accuracy affordable
You cannot verify every record every month, and you do not need to. A random sample of 200-400 records per critical field gives a stable enough accuracy read to spot trends.
Tie thresholds to consequences
A billing address needs a far higher bar than a secondary phone number. Set a target per field based on what breaks when it is wrong, and only escalate when a target is missed twice in a row.
Common questions
What is a good data quality score?
There is no universal number. Set field-level targets from impact — 99%+ for billing and compliance fields, 85-90% for enrichment attributes used in segmentation.
Who should own data quality reporting?
A named owner per critical field, with one person accountable for publishing the monthly summary. Shared ownership reliably becomes no ownership.
Final thoughts
'Our data is bad' is not actionable. Six numbers, tracked monthly against the fields that drive decisions, turn a vague complaint into a work plan.
Need help improving your business data?
DataDabble can research, clean, enrich, and maintain the records your team depends on.
