Testing and validation
See what a data change affects before it reaches production.
A query can run successfully and still lose records or break a downstream report. Embrasure checks proposed dbt changes against production and brings the findings into your pull request.
commerce / analyticsExampleWhat you can check
- Schema changes. Spot removed columns, changed types, and other differences in the output your consumers expect.
- Downstream impact. See which tables could be affected, using confirmed lineage.
- Data differences. Compare row counts, null rates, and distinct values over a recent slice of data.
Keep business-specific tests alongside these checks. For example, matching row counts alone cannot tell you whether a revenue calculation is correct.
Review a change
- Open a pull request in a connected dbt repository.
- Read the findings and check which comparisons completed.
- Investigate unexpected differences, update the change, and review the new results before merging.
A difference may be intentional. Embrasure provides evidence; your team decides whether the result is expected. See the change-review example.
Get started
Connect your dbt repository through the Embrasure GitHub App and add the relevant production BigQuery tables to the catalog. See Change review for setup and supported sources for lineage coverage.
Understand the result
A clean result applies to the checks that ran, the data they covered, and their thresholds. It is not a full row-by-row comparison or proof that two systems produce identical results. Skipped checks and missing lineage remain gaps to review.
For ongoing checks after deployment, use Scouts. For results in GitHub, see Change review.
The proposed output has fewer records. Check whether the join excludes orders you need.