Medallion architecture is an approach, data design pattern¹ or data management framework² used to logically organize data, incrementally and progressively transforming and moving it through three layers: Bronze, Silver and Gold, with each layer progressively transforming raw data into highly refined, analytics-ready datasets.

The idea of moving data through these layers is enriching, improving the structure and quality of data, often performed at data lakehouse platforms.

Medallion architecture is also referred to as "multi-hop" architecture or as Delta Architecture, it is compatible with the concept of a data mesh.

Layers

• Bronze - source data lands on this layer as-is (1:1) with no modifications to structure or data granularity. Either exact representation of source, or/and full history is collected in this layer. Any transformations (aggregation or expansion in row level) are strictly forbidden, except for adding technical columns (to identify ingestion run, facilitate historization).
• Silver - this is a layer where data is cleansed, conformed and structured in a way that data is meaningful and reusable. In other words, integration happens on Silver layer.
• Gold - consumption-ready access layer or interface for users (including business users, analysts, etc.) to specify specific, tailored use cases. This is the layer where data marts sit and special purpose aggregations happen.

See my mapping of Bronze, Silver and Gold to Enterprise Data Platform architecture.

---

1. Medallion Architecture, Databricks Glossary (2026). ↩︎

2. Understanding the Three Layers of Medallion Architecture, ER Studio (2026). ↩︎