Floemdata
Data & ETL

Bronze, Silver, Gold: Data Warehouse Layers

Bronze, Silver and Gold layers organize data from raw form to report. We explain what each layer does and how it appears in Floemdata Analytics.

Floemdata · · 3 min read
Three rising steps with one, two and three marks: bronze, silver and gold data layers

Bronze, Silver, Gold is a common naming scheme in data warehouses: data moves from raw form to report through three layers, getting a little cleaner at each step. Databricks calls this pattern the 'medallion architecture' and describes its goal as improving the structure and quality of data step by step as it flows through each layer.

Bronze: raw and unchanged

Bronze receives data from the source system as it is. According to Databricks, the table structures in this layer correspond to the source system's structures 'as-is', with additional metadata columns. Data written to Bronze is not corrected; it is the raw record you go back to when something goes wrong.

Silver: cleansed and matched

In Silver, data is typed, deduplicated and matched across sources. According to Databricks, the aim is to provide an enterprise view of the key business entities, concepts and transactions.

Gold: ready for reports

Gold holds tables that are ready for reporting and organized for reading; queries need fewer joins.

How it appears in Floemdata Analytics

  • Bronze: append-only. Every extracted version is kept with its extraction time; because values such as cost can change retroactively, the history must be reprocessable.
  • Silver: typed (money, dates, flags) and deduplicated tables; the latest version of each record is selected.
  • Gold: report tables, a calendar dimension and a reconciliation table.

When to build the layers

  • You have several sources and several reports.
  • You have to show where a number came from.
  • For one source and one report, three layers may be too much; start with one and grow.

Source: Databricks, medallion architecture glossary entry.

← All guides