Data Product Factory documentation
The Data Product Factory takes a request for data to a governed table in production. Everybody involved — the team that asks, the owner of the source data, the engineers who build it, the people who audit it — works on one record, and every decision on it carries a name and a time.
A data product goes through three phases. Each has its steps, and each has one gate: a decision that a named person outside the delivery team makes — the source owner approves loading the tables, the source owners and the consumers agree the contract, and the consuming use case signs off the integration test.
Start here
| If you want to | Read |
|---|---|
| See the whole process once, start to finish | Your first data product |
| Know what your job in it is | the guide for your role: use case team, source owner, data engineer, auditor, administrator |
| Run it on your laptop | Run it locally |
| Put it on a server | Deploy |
| Read the record from another system or an AI agent | Read API and MCP, then Connect an agent |
| Understand why it works the way it does | Architecture and Governance |
How this documentation is organised
- Tutorials teach by doing, on a fresh workspace, start to finish.
- Guides by role say what one job decides, where, and what the record keeps.
- How-to guides get one task done: connect Databricks, turn on Entra ID, issue a token, release a version.
- Reference is for looking things up — settings, commands, endpoints, audit codes, permissions, the workflow. Most of it is generated from the code.
- Explanation is the reasoning: the architecture, the governance model, security, and the decision records.
This documentation is versioned with the factory
These pages are part of the application and are deployed with it, so they describe exactly the version that is running — shown at the foot of the sidebar. What changed in each version is in the release notes, and how the version number and these pages are kept in step is in Versioning.