A trust score tells you how far to trust a table, on a scale of 0 to 5 stars. It combines two measures over a recent window of test runs: how well the table's tests pass, weighted by how severe each test is (the quality score), and how many quality dimensions those tests cover (the coverage score).
Validatar shows several numbers with "score" in the name. They are calculated in different ways, over different windows, and some are stored while others are worked out each time the page loads. This article tells you which is which. The detail for each is in the linked articles.
Which score am I looking at?
| Where you see it | Label | Kind | Window | Calculated in |
|---|---|---|---|---|
| Stars at the top of a table page; Trust Score Performance card | Trust Score, Quality Score, Coverage Score | Stored, per table, once a day | Look back period (14 days by default) | How Trust Scores Are Calculated |
| Stars at the top of a schema or data source page; Catalog lists | Avg. Trust Score, Average Quality Score, Average Coverage Score | Stored, averaged across tables | Look back period | How Trust Scores Are Calculated |
| Trust Score Performance chart | Target and Actual | Stored history | 7, 14 or 30 days of daily records | How Trust Scores Are Calculated |
| KPI bar above a test folder or a Catalog Tests tab | Quality Score | Live | The range picked on the bar (24 hours to 365 days) | Live Quality Scores |
| Test Execution Activity panel on a Catalog Info tab | Weighted Score | Live | The range picked on the panel (7, 14 or 30 days) | Live Quality Scores |
| A single test result | Quality Score | Stored, per run | One run | How Trust Scores Are Calculated |
REST API .../trust-scores endpoints |
trustScore, qualityScore, coverageScore | Stored | Look back period | How Trust Scores Are Calculated |
| Data catalogs such as Atlan and Alation | Trust Score, Quality Score, Coverage Score, dimension scores | Copied from the stored score, plus scores the integration works out | Varies by score | Atlan Integration, Alation Integration |
How the stored trust score works
- Every run in the window counts. For each table, Validatar gathers every run of every test linked to the table (or its columns) that started in the look back period. A test that runs daily counts 14 times in a 14-day window.
- Quality score. Each run earns a score from 0 to 1, usually 1 for a pass and 0 for a fail or error, and that score is weighted by the test's severity: Critical 16, High 9, Medium 4, Low 1, Informational 0 by default. The quality score is the weight earned divided by the weight possible.
- Coverage score. Validatar counts the distinct quality dimensions among those runs and looks the count up in the coverage weighting table. By default one dimension already earns 100%.
- Trust score. Quality × coverage × 5.
- Target. Each table's impact rating sets a target trust score. A table at or above its target shows Above Target.
- Roll-up. Schemas and data sources show the plain average of their tables' scores. Included tables with no runs in the window count as 0.
What the live scores are for
The KPI bar and the Test Execution Activity panel answer a different question: how did the tests in this folder or on this object do over the period I picked? They are useful for watching a recent change or a run of failures. They do not include coverage, they do not average tables, and the KPI bar gives no partial credit. Expect them to differ from the stored quality score; Live Quality Scores lists the reasons.
Settings that change the scores
All of these live under Settings > Trust Scores. Users with the Manage Trust Scores permission can override some of them for a data source, schema or table.
| Setting | Affects | Article |
|---|---|---|
| Look Back Period | Which runs count toward the stored scores | Trust Score Settings |
| Coverage Weighting | Coverage score | Trust Score Settings |
| Time Zone | Where each day starts | Trust Score Settings |
| Include or exclude objects | Which tables get a score and enter averages | Trust Score Settings |
| Severity level weights | Stored quality score, KPI bar, Test Execution Activity | Severity Levels |
| Quality dimensions | Coverage score and the dimension breakdowns | Quality Dimensions |
| Impact ratings | Target trust score and badges | Impact Ratings |
On each test, three settings decide its part in the score: its Severity, its Quality Dimension and its Quality Score Method. On each of its metadata links, Impacts Trust Score decides which tables the result counts toward.