Data lineage

Data lineage is a record of a data asset's history: which source systems it came from, what transformations it passed through in a data pipeline, and which downstream tables, reports or models depend on it. It can be captured at the table level or drilled down to individual columns.

Lineage is typically generated automatically from the metadata of transformation tools, dbt builds a lineage graph from its model dependencies, for example, rather than documented by hand, since manual lineage documentation goes stale almost immediately. It complements a data catalog, which describes what an asset is, by showing how it got that way and what breaks if it changes.

Lineage matters most when something goes wrong or needs to change: it lets a team trace a bad number back to its source, or check what will be affected before altering a table's schema. It also supports compliance work under data governance by showing where regulated data flows. Common pitfalls include relying on lineage that only covers part of the stack, for instance transformation logic but not the BI layer, and treating lineage as purely a compliance checkbox rather than an operational tool analysts consult, which also strengthens data observability and data quality.