The short answer
Data lineage is the map of how data moves and transforms: which sources feed which tables, which models depend on which, and which dashboards or jobs consume the result.
Think "who is upstream and who is downstream?"
stripe.charges
-> stg_stripe_charges
-> fct_payments
-> revenue_dashboard
-> finance_close_reportWhy quality teams need it
When revenue_dashboard looks wrong:
- 1Lineage shows blast radius (what else broke).
- 2You walk upstream to find the first bad hop (source vs transform vs join).
- 3You know whom to page (producer owner vs mart owner).
Types people mention
- Table / column lineage: field-level where possible
- Pipeline lineage: Airflow/dbt run graph
- Business lineage: which KPI uses which mart
How you get it
dbt docs/manifest, OpenLineage, catalog tools, warehouse query logs, manual architecture diagrams (fragile alone).
Interview tip: Define lineage as dependency graph for data. Tie it to incident triage: detect → follow lineage upstream → fix → dependents.