A quality dimension names the aspect of data quality a test checks, such as completeness or validity. Each test has one. Dimensions drive the coverage score in the trust score, and they group results in the dimension breakdowns in Projects and the Catalog.
Manage the list on Settings > Trust Scores > Quality Dimensions, where you can add, rename, describe and delete dimensions and see how many tests use each. Set a test's dimension in the test's settings.
Built-in dimensions
| Quality dimension | Question it answers |
|---|---|
| Completeness | Is all the expected data present and complete in the object being tested? |
| Uniqueness | Do the appropriate fields have the uniqueness/cardinality that is expected? |
| Validity | Are the values and combinations of results valid given business rules and expectations? |
| Consistency | Are the data structures, distributions, and values consistent with how they were in the past or in other locations? |
| Accuracy | Are the values accurate to the reality they are designed to represent? |
| Timeliness | Has the data been updated as recently as expected? |
| Standards | Are naming conventions and development standards being followed? |
| Relationships | Are the defined relationships between objects being consistently followed? |
| Security | Are the appropriate security and access rules defined and applied correctly? |
| Precision | Is the level of detail and data type precision consistent with expectations? |
| Performance | Are the systems responding in a timely manner and performing as expected? |
| Processing | Are all the jobs/processes/pipelines running and completing as expected? |
| Compliance | Are applicable policies and compliance rules being applied? |
| None | This test has not been assigned a quality dimension. |
How dimensions affect the trust score
Dimensions affect the coverage score only. They do not change the quality score.
For each table, Validatar counts the distinct dimensions among the test runs counted in the look back period, then reads the coverage score for that count from the Coverage Weighting table (see Trust Score Settings).
- A dimension counts once any test in it ran against the table in the window, whether it passed, failed or errored.
- Tests with the Informational severity count. They have no effect on the quality score, but they still add their dimension to coverage.
- None never counts. A table whose tests all have the None dimension has a coverage score of 0, and so a trust score of 0, however well its tests pass.
- Which dimension a run counts under is fixed when it runs. Changing a test's dimension affects later runs only.
- All dimensions count the same. Covering Completeness and Validity earns the same coverage as covering Standards and Security.
Tip: Give every test that should count toward trust a real dimension. Tests left on None are the most common reason a well-tested table shows a trust score of 0.
Dimension breakdowns
Several views split results by dimension. They are calculated differently:
| View | What each dimension shows |
|---|---|
| Test Execution Activity panel, Quality Dimension table | Severity-weighted results of the runs in the date range |
| KPI bar, Quality Dimension mode | Share of runs in the date range that passed, unweighted |
| Data catalog integrations such as Atlan and Alation | Share of the table's tests whose latest run passed. See Atlan Integration and Alation Integration |
None of these is a per-dimension trust score. Validatar does not store a score per dimension.
Each dimension also carries a built-in weight, which is not shown in settings. It does not affect the trust score or the coverage score; see Live Quality Scores for the one place it shows.