Data quality · Extra High-Value

What is a quality gate?

Mediumquality gateCIdbt buildpublishseverity

Interview question

What is a data quality gate?

How to answer

The short answer

A quality gate is a checkpoint that data (or a pipeline stage) must pass before it is promoted to the next environment or published to consumers. Fail the gate → block merge, block publish, or mark the asset unhealthy.

Output
raw load -> staging tests (gate 1) -> mart build -> mart tests (gate 2) -> BI publish
                 |                                    |
              fail: stop                           fail: do not certify

Where gates appear

  • CI: contract/schema tests on producer PRs
  • Orchestration: Airflow task fails if GX/dbt tests fail (dbt build)
  • Warehouse: only "certified" tables appear in the trusted catalog layer
  • ML: training job refuses features failing freshness/completeness gates

Gate design tips

  • Separate warn vs block severities
  • Gate on high-impact assets first (revenue, billing)
  • Make failures actionable (which check, which , who owns it)

Interview tip: "A quality gate is a pass/fail checkpoint before publish." Mention dbt build, CI contract tests, and warn vs error.