Skip to main content

Glossary

What is medallion architecture?

Last updated Monday, Aug 3, 2026

Medallion architecture organizes a data platform into three layers, Raw, Clean, and Ready (also known by the traditional names bronze, silver, and gold), that take source data and refine it step by step into trusted, business-ready tables. Each layer has a single, fixed job, so a schema change or a bad transform in one layer doesn't quietly break the others.

The three layers

  • Raw (bronze): Data lands here exactly as the source sent it. Zero transforms, full history, stored with Iceberg ACID guarantees so nothing is lost and nothing is silently overwritten.
  • Clean (silver): LLM-generated SQL cleans, validates, and joins the Raw data. Every change is user-approved and version controlled before it runs.
  • Ready (gold): Business metrics and a star schema built on top of Clean, updated incrementally so dashboards and reports stay current without a full rebuild.

How OptimaFlo implements this

The Data Engineer role builds the pipelines that move data through Raw, Clean, and Ready, landing generated SQL on a canvas for you to review before it runs. The Analytics Engineer role writes and re-runs the SQL behind Clean-layer models, checking column references and types before anything executes.

For a deeper walkthrough of why each layer exists and what breaks without it, see the medallion architecture explained.

Related terms

  • AI data team: The roles, including the Data Engineer and Analytics Engineer, that build and maintain a medallion pipeline.
  • BYOC: Where the Raw, Clean, and Ready tables actually live: your own cloud account, not a vendor's.
  • Apache Iceberg: The open table format Raw, Clean, and Ready are stored in.
  • Bronze, silver, and gold: The traditional names for these same three layers.
  • Data transformation: The reshaping work the Clean layer's SQL does, run against data already landed in Raw.
  • Star schema: The model the Ready layer typically builds: a fact table joined to dimension tables for BI queries.

Frequently asked questions

Each layer has one job. Raw preserves the source exactly as it arrived, so you can always reprocess if a downstream rule turns out wrong. Clean applies validated logic once. Ready pre-aggregates the metrics people query most. Collapsing them into one table means every fix risks breaking every other use of that data.

Yes, same three layers, different names. Bronze, silver, and gold are the traditional terms for this pattern. OptimaFlo uses Raw, Clean, and Ready for user-facing docs and the product UI, and keeps bronze, silver, and gold as internal table prefixes and catalog names.

No. The Data Engineer role builds the pipelines that move data through the layers and lands the generated SQL on a canvas for you to review. The Analytics Engineer writes and re-runs the SQL behind Clean-layer models, and you see that generated SQL before anything runs.

Browse every term or see the AI data team roles.

Staffed, not self-serve

See AI teammates work on your own data.

Start your pilotContact

Now in early beta. One flat plan, no per-query tax. Runs in your cloud, and your tables stay in your account.