Databricks — What is Medallion Architecture

Databricks' canonical description of the medallion (multi-hop) lakehouse pattern. Agent-fetched 2026-06-26 to ground the Medallion Architecture entity.

Summary

Databricks defines the medallion architecture as a data design pattern that incrementally and progressively improves the structure and quality of data as it flows through layers — bronze → silver → gold:

• Bronze (raw): the entry point where source data lands unchanged; optimized for change-data-capture, historical archive, lineage, and auditability. Minimal validation; permissive types to survive schema drift.
• Silver (cleansed & conformed): "just-enough" refinement into an enterprise view — cleansing, deduplication, normalization; at least one validated, non-aggregated representation of each record.
• Gold (business-level): consumption-ready, de-normalized, read-optimized models for business users.

Databricks frames it as building "a single source of truth for enterprise data products" with ACID guarantees and the ability to recreate tables from raw.

Key claims

• The layers denote data quality tiers, not just storage stages.
• Bronze is explicitly where data lands; the modeling happens above it — the vocabulary behind the wiki's Lands vs Lives distinction.
• The doc does not attribute an individual coiner; Databricks is the popularizer.

Connections

• Medallion Architecture — the entity this source grounds (raw landing → cleaned → business layers).
• Lands vs Lives — "bronze is where data lands" is the literal source of the lands-vs-lives framing; in a view-only medallion the upper layers are views.
• One Fact Per Grain — the gold layer is where dimensional modeling (grain) lives.
• System of Record — Databricks pitches the pattern as a "single source of truth."

Source

• Original: Databricks, "What is Medallion Architecture?" (source-url above); faithful capture in source-file. Living vendor documentation — no fixed date.