Momentum One Academy · glossary
Medallion Architecture
Definition
Medallion Architecture is a lakehouse design pattern that separates data into three quality stages: bronze retains source data in its arriving form, silver validates, standardizes, and enriches it, and gold shapes trusted data for reporting and other consumption. The pattern preserves a traceable source of truth while transformations progressively improve analytical usability. Microsoft identifies it as a recommended design approach for Fabric. 1 3
Where it sits in Fabric
In Fabric, the pattern is implemented over OneLake, the tenant-wide logical data lake. A layer can be a separate lakehouse, or a bronze/silver lakehouse can feed a gold warehouse; the right topology depends on workload and team needs. Lakehouses provide Files and Tables, Spark-based engineering, and a SQL analytics endpoint, while Power BI and other consumers read the refined outputs. Delta Lake is Fabric’s default table format and is typically used for silver and gold tables, giving engines a common, transactional representation. 1 2 4 5
When to use it
Use this pattern when raw inputs need repeatable cleansing, quality checks, and business modeling before broad analytical use. It is especially useful when engineering and BI teams share data across Spark, SQL, pipelines, and Power BI without creating separate copies. Keep bronze close to the source, make silver the trusted integration layer, and design gold around stable business questions rather than around every upstream system. 1 2
Practical example
A retailer lands JSON web events, CSV inventory extracts, and CRM exports in a bronze lakehouse. Silver jobs standardize timestamps, remove duplicate events, validate product keys, and convert currencies. Gold tables then expose daily sales, inventory position, and customer lifetime value for a Power BI semantic model through the lakehouse SQL analytics endpoint. 1 3
Production caution
Layering alone does not guarantee performance or recoverability. Delta tables need an explicit maintenance plan because excessive small files can slow Spark, SQL analytics, Direct Lake, and external readers. Schedule compatible compaction and cleanup, and do not shorten VACUUM retention until time-travel requirements and concurrent readers or writers have been assessed. 1 6
Workshop connections
In Module 1, position Medallion Architecture within Fabric’s OneLake and workspace model. In Module 2, connect the bronze, silver, and gold stages to lakehouse storage, Delta tables, and Spark/SQL access. In Module 4, apply the pattern to ingestion, transformation, validation, and curated consumption design.
Ready to turn these layers into a governed delivery pattern? Explore the Momentum One Academy Fabric Production Workshop.
References
Plan your workshop