Lineage
What feeds what, and what is affected when something upstream breaks.
Data Warehouse → Lineage is a graph of which tables feed which.
The two questions
- This table is stale — what else is now wrong?
- This number looks wrong — where does it come from?
Both are asked under pressure, and neither has an answer from a list of tables.
Building it
Lineage is derived from query definitions: a scheduled query, alert query or dashboard tile that reads table A and writes table B creates an edge.
Edges can also be declared for pipelines AccelerUp does not run.
Derived edges are only as complete as what AccelerUp can see. A transformation run entirely outside — dbt, an Airflow DAG — is invisible unless declared. The graph shows which edges were discovered and which were declared, so you can tell how much of the picture you are looking at.
Using it with quality checks
Lineage plus data quality gives you impact analysis: a freshness check failing on a source table highlights everything downstream that is now reading stale data — including the dashboards and the alerts.
That last part matters. A stale table does not only make a dashboard wrong; it makes every alert reading it wrong in the reassuring direction.
Related
- Dependency map — the same idea for services
Where this behaviour lives: frontend/src/app/dwh/lineage/. If the code and this page disagree, the code is right — please fix the page.
Part of Data warehouse — Querying the warehouse, and alerting on what the query returns.