Skip to main content
Validation compares a Job’s source and destination across accuracy dimensions. It is broader than the automatic post-sync schema validation row-count check and lives on the Job’s Data Quality tab.
This is not Trust → Data Quality (catalog test cases). Pipeline validation audits whether the sync was faithful. Catalog DQ tests assert business rules on inventory assets.

​
Accuracy dimensions

​
Run validation

  1. Data → Jobs → open the Job.
  2. Open the Data Quality tab.
  3. Click Run Validation.
Empty state before the first run: Not yet validated. Options (mode, sample percent, timeout, table subset) follow the dialog / API fields exposed in the UI (full vs sample, and so on).

​
Semantic rules

On the same Data Quality tab:
  1. Open New semantic rule (add rule control).
  2. Enter a name and a SQL expression that returns rows only when data violates the rule.
  3. Save. Rules re-evaluate on each validation run.
Use portable SQL (for example COUNT(*)-style checks) that can run on both source and destination.

​
Validation vs Trust Data Quality

Last modified on October 10, 2026