Learn
What is Medallion Architecture?
A layered approach to organising data: raw data lands first (bronze), gets cleaned (silver), then shaped into business-ready tables (gold) – so your team always knows which numbers are trustworthy.
What it is
Medallion architecture is a way of organising data in layers, like refining crude oil into usable fuel. The bronze layer stores raw data exactly as it arrives from your source systems – warts and all. The silver layer cleans it: fixing formats, removing duplicates, standardising names. The gold layer shapes it into the tables your dashboards and reports actually use. Think of it like a commercial kitchen. Deliveries arrive at the loading dock (bronze) – unpackaged, unchecked. The prep team cleans, portions, and stores ingredients properly (silver). The chefs assemble finished dishes ready to serve (gold). Nobody serves raw deliveries to customers.
Why it matters for your business
Without layers, your data warehouse becomes a tangle where nobody knows which table is the "right" one. Teams build their own versions of revenue, their own customer counts, and their own margin calculations – defeating the purpose of centralisation. A $55M distribution company we worked with had 340 tables in their warehouse with no naming convention or layering. It took analysts 30 minutes just to find the right table, and they still were not sure if the numbers matched finance. After restructuring into bronze-silver-gold layers, every analyst knew exactly where to go: gold for trusted business metrics, silver if they needed more granular data, bronze if they needed to audit the raw source.
How we approach it
We implement medallion architecture using tools like dbt to define each transformation as version-controlled code. Bronze tables are ingested automatically via pipeline tools. Silver transformations handle deduplication, type casting, and standardisation. Gold tables are modelled around business concepts – "monthly revenue by product line" rather than raw transaction rows. Every promotion from one layer to the next has quality checks built in, so problems surface early rather than in a board meeting.
Related Services
Key Takeaways
- •Bronze-silver-gold layers give your team one clear place to look for trusted data – no more guessing which table is right.
- •Raw data is always preserved in bronze, so you can rebuild or redefine calculations without going back to source systems.
- •Quality checks at each layer transition catch problems before they reach your dashboards.
- •This pattern works with any modern warehouse – BigQuery, Snowflake, Databricks, or Postgres.