Bronze Ingestion

Summary

Bronze is the lakehouse layer for raw or source-shaped ingested data. It keeps external system facts available before the platform normalizes them into canonical internal models.

Problem

External systems change shape, expose different natural keys, and carry provider-specific semantics. If the platform tries to canonicalize everything directly at the edge, it loses useful evidence and makes backfills harder.

Principle

Ingest source data into bronze first. Preserve enough source shape to audit, rebuild, and reprocess while keeping provider-specific mess from leaking into internal services.

Platform Shape

The lakehouse docs place ingestion under pipelines/lakehouse/ingest/. ServiceTitan, QuickBooks Online, Spectrum, and Microsoft ingestion flows write source records into bronze schemas or source-specific bronze areas. Jobs under jobs/lakehouse/ orchestrate those pipeline runs and record execution through the job runner.

Bronze is not the serving layer for business workflows. It is the evidence layer that later silver and gold transforms use.

Source Evidence

  • docs/domains/lakehouse.md
  • pipelines/lakehouse/ingest
  • jobs/lakehouse
  • docs/engineering/concepts/silver-canonicalization.md
  • docs/engineering/architecture/lakehouse-service-boundary.md