Lineage you can explain at the whiteboard

Lightweight lineage practices for ETL teams that cannot staff a full catalog program — and still need to answer ‘where did this number come from?’

Enterprise catalog programs stall when they demand perfect coverage before useful coverage. Pipeline teams still need to answer a basic question: which jobs touch this mart column?

Start from the questions that hurt

Collect the last ten times someone asked “where did this number come from?” Map only those paths first. A partial lineage graph that answers real questions beats a yellow status bar in a tool nobody opens.

Keep lineage next to the job

A markdown table in the same repository as the transform — source system, extract job, transform job, load target, owner — ages better than a wiki page orphaned from the code. Pull-request templates can require an update when columns change.

When to deepen

Once the painful paths are documented, invest in automated lineage for high-churn areas. Leave stable reference data to manual notes until churn justifies the machinery.

How we use this in assessments

In findings documents we often include a one-page lineage sketch for the critical feed under review. Clients tell us that sketch gets photocopied onto desk whiteboards more often than our longer appendices.