Trust Scores
Trust Scores at /data-quality/trust-scores give every table a 0-100 health rating. It's the table-level granular companion to the DQ Scorecard's domain-level rollup.
Composition
Each table's score is a weighted combination of:
| Dimension | Default weight | Source |
|---|---|---|
| Rule pass rate | 35% | DQ Results |
| Freshness | 20% | Data Freshness |
| Volume stability | 15% | Volume Monitor |
| Schema stability | 10% | Schema Drift |
| Anomaly density | 10% | Anomalies |
| Certification status | 10% | Certifications |
Weights are configurable in clxs.yaml under data_quality.trust_scores.
What the page shows
- Top-trusted tables — sorted by score descending
- Watchlist — tables below their target threshold
- Recently regressed — biggest drops in last 7 days
- Recently improved — biggest gains
Each row has the score, a colour-coded badge, and a sparkline of last-30-day trend.
Targets
Set per-table or per-domain target scores in Targets sub-tab. Below-target tables fire alerts via Alert Rules.
Drill-in
Clicking a table opens a panel showing each dimension's current score, weight, and trend. This is where you diagnose why a table's trust score dropped — was it freshness? Schema drift? A new failing rule?
Stale-data adjustment
Tables with no rule executions get a default score (50) with a no_coverage flag. The fix is to either add rules or mark the table as not_in_scope to exclude it from scoring.
API
GET /data-quality/trust-scores
GET /data-quality/trust-scores/{table_fqn}
GET /data-quality/trust-scores/{table_fqn}/history
PATCH /data-quality/trust-scores/{table_fqn}/target
Related
- DQ Scorecard — domain-level rollup
- Coverage Map — which tables have rules
- Watchlist alerts