Data Lineage

Data lineage is the map of where data comes from and where it goes: which source produced a value, which transformations touched it, and which downstream tables, dashboards, or models depend on it. It answers two crucial questions — "where did this number come from?" (upstream) and "what breaks if I change this column?" (downstream). For SaaS builders, lineage is what turns debugging a wrong dashboard from a day of Slack archaeology into a five-minute trace. When a KPI looks off, you follow the lineage back through each pipeline step to find where it diverged. Before a schema migration, you check downstream lineage to see who you'll break. Modern tools capture lineage automatically. dbt builds a dependency graph from your SQL models; column-level lineage tools (in DataHub, OpenMetadata, or warehouse-native features) parse queries to trace individual fields. Practical note: column-level lineage is far more useful than table-level for impact analysis, but harder to maintain — invest in it for your most business-critical metrics first, and let automated parsing do the heavy lifting rather than manual diagrams.