What is Medallion Architecture?
A medallion architecture is a way of organizing data in a lakehouse into three progressive layers — Bronze (raw), Silver (cleaned and conformed), and Gold (business-ready) — so data gets more trustworthy and useful as it moves up each tier.
At a glance
- Three layers: Bronze (raw ingest), Silver (cleaned/joined), Gold (aggregated, business-ready).
- Each layer is a checkpoint — quality and structure improve as data moves up.
- Popular in lakehouses (Databricks, dbt/SQLMesh) because each layer is a queryable table.
- Makes pipelines debuggable: you can inspect data at every stage.
What each layer does
Bronze is the landing zone: raw data copied from source systems with little or no transformation, kept as a faithful historical record. Silver cleans, deduplicates, and joins that raw data into conformed, validated tables that model real business entities. Gold aggregates Silver into the specific tables that dashboards, reports, and ML features read from.
The value is the progression. Instead of one giant, opaque transformation from source to dashboard, you get named checkpoints. When a number looks wrong, you can trace it layer by layer — a huge win for trust and debuggability.
Why teams adopt it
A medallion architecture gives a messy pile of source data a clear, repeatable shape. Bronze preserves the source of truth; Silver becomes the reliable, reusable core; Gold serves fast, purpose-built outputs. New use cases usually just add a new Gold table on top of existing Silver, rather than re-plumbing everything.
It pairs naturally with transformation tools like dbt or SQLMesh (which define each layer as SQL models) and with change-data-capture pipelines that keep Bronze continuously up to date.
Frequently asked questions
What are the three layers of a medallion architecture?
Bronze (raw, unmodified source data), Silver (cleaned, deduplicated, and joined into conformed tables), and Gold (aggregated, business-ready tables for dashboards, reports, and ML).
Is medallion architecture the same as ETL?
It's a way of organizing the output of your ETL/ELT into layered tables. ETL/ELT is the movement and transformation of data; the medallion pattern is how you structure the results into Bronze/Silver/Gold checkpoints.
Do I need Databricks to use a medallion architecture?
No. It's a design pattern, not a product. You can implement Bronze/Silver/Gold on Databricks, Snowflake, BigQuery, or any warehouse/lakehouse, often using dbt or SQLMesh to define the layers.
Why use three layers instead of transforming data once?
The layers act as checkpoints. They make pipelines easier to debug, let multiple Gold outputs reuse the same trusted Silver core, and preserve raw history in Bronze so you can reprocess if logic changes.
Putting Medallion Architecture to work?
We help data & AI teams design and ship this in production. Tell us what you're building and we'll point you at the shortest path.
Talk to our team