Discovered by downstream consumers when reports look wrong

DQ is what someone notices in the dashboard. No proactive checks.

Periodic manual sampling by analysts before major deliverables

Spot checks before reports go out. Cadence-driven, not continuous.

Automated DQ checks at ingest with named KPIs (completeness, freshness)

DQ violations page on-call; thresholds defined per dataset.

Automated DQ + named owners per dataset; SLAs enforced

Datasets have explicit owners with documented quality contracts.

Data contracts with producers + consumers, automatically tested in CI

Schema + semantic contracts pre-tested before changes ship.