Data Lineage

A practical example

Example: when a revenue column's logic changes, column-level lineage should show every downstream metric, report and alert affected, so the team can notify owners before numbers shift.

What to evaluate before investing

• Ask whether lineage is parsed from code like SQL and dbt, inferred from query logs, or declared manually, and how each mode handles dynamic SQL.
• Verify column-level lineage exists for your transformation layer, not just table-level edges between systems.
• Test how lineage updates when pipelines change: delayed or stale graphs undermine trust during incidents.

Limitations and tradeoffs

Lineage graphs are only as complete as their capture method; log-based approaches miss logic outside queries, and manual approaches go stale, so coverage claims deserve scrutiny.