Learn
Data quality checks worth running every time
Five dimensions catch nearly every data problem that reaches a report. Run them before analysis, not after someone questions a number.
Updated September 20, 2026
The five checks
| Dimension | Question | Detection |
|---|---|---|
| Completeness | Are rows or periods missing? | Row counts vs source, gaps in the date series |
| Validity | Do values fit their type and range? | Type profiling, min and max, negative checks |
| Uniqueness | Are keys unique? | Duplicate count on the key columns |
| Consistency | Do categories and totals agree? | Unique value lists, reconciliation to a known total |
| Timeliness | Is the data current enough? | Max date vs expected refresh |
Make the checks cheap
A profiling table that shows type, blanks, uniqueness and ranges per column answers four of the five checks at a glance. Generating it should take seconds, otherwise it gets skipped.
Frequently asked questions
How do I check quality quickly?
Load the file into the spreadsheet analyzer to get a per-column profile, then reconcile one total against the source system.
Related reading
Analyze your data with AI
Open the GridMind workspace, bring in a CSV, Excel file or your Power BI model, and let AI build the formulas, charts and dashboards for you.