All case studies
Case Study · 2025 · Internal R&D · 9 engineering days

Warehouse Lineage Layer

Column-level lineage and freshness monitoring across an ELT estate feeding executive reporting.

Reference implementation for our analytics-trust standard.

warehouse.internal / lineage / dim_customerLineageModelsLineageColumnsTestsFreshnessDocsField-level lineage — dim_customerreverse ETL · 218 columns traced · 0 orphaned fieldsGraphsalesforce.accountsource · apistg_accountstaging · viewdim_customermart · tableColumn traceCOLUMNDEPTHcustomer_id3 hopslegal_name3 hopsregion_code4 hopslifetime_value5 hopschurn_flag5 hopsupdated_at2 hopssource_system1 hop
The problem

Once a reporting estate passes a few dozen models, nobody can answer two basic questions with confidence: where did this number come from, and is it current? Dashboards get quietly distrusted, and teams rebuild the same metric in three places.

What we built
  • A parser that derives column-level lineage from warehouse query history and model definitions, without requiring analysts to hand-maintain a catalogue.
  • Freshness and volume monitors per model with expected-window alerting, so a silently stalled load surfaces before the morning review.
  • An impact view that answers, for any source column, which downstream models and dashboards break if it changes.
  • A metric registry giving each executive figure one owning definition, with drift detection when a second definition appears.
The result

On an estate of 140 models and 900 columns, lineage extraction achieved full coverage of parsed models and cut the time to answer a where-did-this-number-come-from question from open-ended investigation to under a minute.

Want this level of documentation on your project?

Every engagement ships with the same written record: what changed, why, and what it measured.

Start a conversation