Analysis models
An analysis model is the contract between your database and everyone who uses the data.
- Tables and relationships on a visual canvas, including SQL views and star-schema tidying.
- Hierarchies and time dimensions that drive drill-down and period comparisons.
- Measures and calculated measures with units and display scale, so a number reads the same in every chart.
- Business semantics for AI: names, descriptions, and synonyms that help the AI Agent map a question to the right fields.
- Model diagnostics in the Insights panel flag problems before they reach a dashboard.
The Metrics Library
The Metrics Library stores governed metric definitions with stable IDs and a Draft or Certified status.
- Bind measures to enterprise metrics. Models declare which certified metric a measure implements.
- Catch drift. An AI-assisted Compare definition check flags measures whose logic no longer matches the certified definition. Diagnostics flag duplicate or stale bindings.
- Generate drafts from existing models. Since 10.00, draft metrics can be generated from a model’s measures. Certifying a metric requires a definition and a unit.
- Import and export definitions as JSON or CSV.
Why it matters
Without a shared model, every report, customer portal, and AI tool re-implements the same logic, and the numbers drift apart. With Datafor, the definition lives in one place and every surface inherits it, including the permissions attached to it.