What Is Medallion Architecture?
Medallion architecture is a data design pattern that organizes a data platform into a series of layers, each one refining the data a step further. The layers are conventionally named bronze, silver, and gold, after medal tiers, to signal increasing value and quality. Raw data lands in bronze, gets cleaned and conformed in silver, and is shaped into business-ready models in gold. The pattern became popular with the lakehouse, where it gives an otherwise open and flexible storage layer a clear, governed structure.
The idea is simple but useful: rather than transforming raw source data straight into finished reports in one opaque step, medallion architecture breaks the journey into stages that are each understandable, testable, and reusable. Every layer has a defined purpose, and data only moves forward once it meets that layer’s standard.
The value of the layered model is not the three names. It is that each stage has one job and a clear definition of done, which tends to make a pipeline easier to trust and to debug when something looks wrong.
Marla Nelson, CTO
The Three Layers
Bronze: Raw, As-Ingested Data
The bronze layer holds data exactly as it arrives from source systems, with little or no transformation. It is the landing zone: ERP tables, application exports, files, and streams captured in their original form. Keeping a faithful copy of the raw data matters, because it means the pipeline can be rerun or corrected later without going back to the source. Bronze is usually append-only and carries metadata about when and where each record came from.
Silver: Cleaned and Conformed
The silver layer is where data is validated, cleaned, and integrated. Duplicates are removed, types are standardized, codes are decoded, and data from different sources is joined into consistent, conformed tables. A customer record scattered across three systems becomes one reconciled customer in silver. The result is an enterprise view of the data that is correct and consistent, though not yet shaped for any particular report.
Gold: Business-Ready Models
The gold layer is the data shaped for consumption: aggregated, modeled, and organized around business needs. Star schemas, metrics, and subject-area models live here, ready for reporting, dashboards, and analytics. Gold tables are what business users and BI tools actually query. Because the cleaning and conforming already happened in silver, gold can focus on presentation and performance.
How Data Moves Through the Layers
Data flows in one direction, bronze to silver to gold, and is refined at each step. Ingestion lands raw data in bronze. Transformation jobs read bronze, apply cleaning and conforming rules, and write silver. Further modeling reads silver and builds the aggregated, analytics-ready tables in gold. Each step is a distinct, observable stage with its own logic and its own quality checks, rather than one monolithic transformation that is hard to reason about.
Because each layer is persisted, the pipeline is restartable and auditable. If a gold model looks wrong, an engineer can trace the issue back through silver to bronze and find exactly where it entered, instead of re-running an opaque end-to-end process and hoping.
Why Use a Layered Approach
The layered pattern brings several practical benefits:
- Debuggability: problems can be isolated to a layer, since each has a clear definition of correct.
- Reusability: many gold models can be built from the same well-conformed silver, instead of each report re-cleaning the raw data.
- Auditability: keeping raw bronze data means the lineage from source to report is traceable.
- Incremental processing: layers can be refreshed independently and incrementally, rather than rebuilding everything each time.
- Separation of concerns: cleaning logic lives in silver, presentation logic in gold, so changes stay contained.
Together these make a platform easier to maintain and trust as it grows, which is why the pattern scales well from a single team to an enterprise.
Medallion Architecture and the Lakehouse
Medallion architecture is closely associated with the data lakehouse, and the two fit together naturally. A lakehouse stores all data in open formats on inexpensive cloud storage while adding the reliability and governance of a warehouse. Medallion architecture gives that storage a coherent structure: bronze, silver, and gold become organized zones within the lakehouse, each governed and queryable. The pattern is what keeps a lakehouse from becoming an unmanaged data swamp.
Medallion vs Traditional Staging
The idea of staging data in steps is not new; data warehouses have long used staging and presentation layers. What medallion architecture adds is a clear, named convention and the expectation that every layer is a persisted, queryable, governed asset rather than a throwaway staging table. In a lakehouse, all three layers are open and accessible, so silver is not a hidden interim step but a genuine enterprise resource that many gold models, and even some direct analyses, can build on.
Common Pitfalls
The pattern is a guideline, not a rulebook, and it can be applied poorly. A few common mistakes:
- Treating the names as the goal, creating three layers out of habit without a clear purpose for each.
- Letting silver do too little, so cleaning logic leaks into many gold models and gets duplicated.
- Skipping governance, so the layers exist but data quality and definitions are not actually enforced.
- Over-engineering small pipelines that do not need three full layers.
The benefit of the layers comes from a clear contract at each stage. When that contract is vague, the structure becomes overhead rather than help.
Medallion Architecture in a Governed Foundation
For companies bringing ERP data into analytics, medallion architecture maps cleanly onto the work that has to happen: land the raw JD Edwards, Vista, NetSuite, or OneStream data, clean and conform it into an enterprise view, then shape it into models ready for Power BI. QuickLaunch builds that layered, governed foundation so teams do not have to design and maintain it themselves. Raw source data is captured, conformed into consistent enterprise tables, and modeled into report-ready structures, with the governance and lineage that make each layer trustworthy.
The result is the benefit the pattern promises, a pipeline that is dependable, traceable, and reusable, without the months of platform engineering it usually takes to build one by hand.
Frequently Asked Questions
What is medallion architecture?
A data design pattern that refines data through three progressive layers, bronze (raw), silver (cleaned and conformed), and gold (business-ready). Each layer improves quality and readiness, and data flows in one direction from bronze to gold.
What are the bronze, silver, and gold layers?
Bronze holds raw data as ingested from source systems. Silver holds cleaned, validated, and integrated data in a consistent enterprise form. Gold holds aggregated, modeled, business-ready tables that reporting and analytics query directly.
Why is medallion architecture used with the lakehouse?
The lakehouse provides open, low-cost storage with warehouse-like governance, and medallion architecture gives that storage a clear structure. The layers turn an otherwise flexible store into governed, queryable zones, which keeps a lakehouse from becoming a data swamp.
