Declare source freshness SLAs and make stale data fail before downstream work
When stakeholders notice yesterday's data before the data engineer does.
- Data quality
Larger outcome
4 RecipesData reliability
Recipes whose deliverable is confidence rather than a table: freshness that fails before downstream work runs, gating on readiness instead of on a clock, and proof that a change is safe.
Gate a consumer pipeline on upstream data readiness instead of a clock
When the mart runs at a fixed hour whether or not the upstream load actually finished.
- Orchestration
Prove an existing dbt change is safe to merge without modifying it
When a pull request changes dbt models and nobody wants to merge it on faith.
- Data quality
Reconcile equivalent data across two live platforms before cutover
When two platforms are running the same logic in parallel and nobody has proven they agree.
- Data quality