Lakehouse Publication State Machine

Summary

Lakehouse publication turns source data into trusted operational products. Its state machine is how the platform keeps ingestion, canonicalization, analytical facts, serving snapshots, and UI consumption distinct.

Reader Question

Which states does data pass through before a UI or workflow can safely consume it, and how are stale or failed publications represented?

State Machine

Source Captured records external evidence. Bronze Loaded preserves source shape. Silver Canonicalized resolves entities and relationships. Gold Built creates domain facts. Quality Checked decides whether the product is safe to publish.

Published is an explicit transition, not an accidental query path. It creates serving tables or snapshots for UI, API, review, or downstream jobs. Serving means consumers have a stable projection. Stale or Failed states make freshness visible and prevent silent dependence on incomplete analytics.

This state machine is the conceptual bridge to a future physical OLTP and analytics split. The platform can use one database today only because ownership and publish transitions remain explicit.

Visual

The diagram shows the rebuildable data path and the explicit publish boundary before serving state.

Related Threads

• Data Boundaries
• Lakehouse And Execution
• Data Publication Methodology