One metric per dimension
Data quality is measured through six dimensions. For each important data element (for example, a customer email), one simple metric per dimension is enough, expressed as the percentage of records that meet the rule.
| Dimension | Metric | Example |
|---|---|---|
| Completeness | % of records with the field filled in | Customers with an email address |
| Accuracy | % of values matching the trusted source | Addresses matching the official register |
| Consistency | % of records that agree across two systems | Same customer in CRM and billing |
| Timeliness | % of data updated within the deadline | Prices reviewed this month |
| Validity | % of values with the right format and range | Five-digit postal codes |
| Uniqueness | % of records without a duplicate | Customers with no repeated record |
How many metrics you need
Fewer than it seems. Pick the few data elements that matter most to the business, apply the metrics that make sense for each one, and leave the rest until that works. Measuring hundreds of fields that nobody acts on improves nothing.
How to use them
- Set a threshold: the minimum acceptable level is decided by whoever uses the data, not by the technical team.
- Measure the same way every time: same rule and same cut-off date, so results are comparable.
- Assign an owner: a metric without an owner does not get fixed. See who should own data quality.
For a first indicative figure, try the free data quality calculator. For KPI and dashboard detail, read how to measure data quality.
Quick answers
How many data quality metrics do you need?
One per dimension and per critical data element. Start with a few important data elements and expand once those metrics are being used to take action.
How do you calculate a data quality metric?
Divide the number of records that meet the rule by the total number of records and express it as a percentage.
Who sets the acceptable threshold?
Whoever uses the data to make decisions, usually the business area owner, not the technical team.