Medallion Architecture

Concepts: dmMedallion

Bronze, Silver, Gold: Progressive Refinement Medallion architecture organizes a data lake into three layers, each with a clear purpose. Raw data enters at bronze. Cleaned and validated data lives in silver. Business-ready aggregates and curated datasets live in gold. The architectural principle is separation of concerns: raw ingestion should not depend on business logic, and business transformations should not depend on source system quirks. This is the most common architecture at companies using Databricks, Delta Lake, or any modern lakehouse platform. If you work with a data lake, you will encounter this pattern. Bronze is append-only and never cleaned. This is intentional. If your silver transformation has a bug, you can fix it and reprocess from bronze. If bronze was already cleaned, y