Medallion architecture describes data readiness and quality progression across storage layers, while Kimball dimensional modeling provides a structured technique for organizing business facts and dimensions for analytical queries. They are complementary concepts: medallion defines the refinement pipeline, and Kimball modeling typically structures the final business-ready gold layer.
Medallion quality stages
The medallion pattern organizes a modern data lakehouse into three distinct refinement tiers:
- Bronze: Raw, append-only landing zone storing source system payloads exactly as received, often in Parquet or Delta format with raw JSON and ingestion metadata.
- Silver: Cleaned, validated, and conformed data where types are cast, duplicates removed, and entities standardized into normalized or 3NF structures.
- Gold: Business-level presentation layer optimized for reporting and analytics, designed to answer specific business questions.
Where Kimball fits in gold
Kimball dimensional modeling finds its natural home in the gold tier. Fact tables like fact_monthly_revenue and dimension tables like dim_customer are built by consuming cleansed silver entities. Conformed dimensions and surrogate keys are established here so self-service BI tools can execute clean star-schema joins without data quality issues.
Comparing the two frameworks
Thinking of medallion and Kimball as competitors is a common beginner misconception. Medallion dictates how raw data matures through verification stages, while Kimball dictates how clean data is modeled for human understanding and fast SQL slicing. A well-designed pipeline uses medallion engineering to produce Kimball data marts.