Model Health

The Model Health dashboard gives you a comprehensive view of your semantic model's quality. It scores your model across multiple dimensions, identifies issues, and provides actionable findings to help you improve your model before publishing.

Why Model Health Matters for AI

As AI tools like Copilot, MCP servers, and natural-language query engines become a primary way people interact with semantic models, model quality directly affects how well AI can understand and work with your data. A model with missing descriptions, ambiguous naming, bidirectional relationships, or no date table forces AI to guess - and guessing leads to wrong answers.

The AI Readiness score measures exactly this: how well your model is structured for AI consumption. Every factor it tracks - description coverage, synonym coverage, relationship simplicity, star schema adherence - maps to something AI agents rely on to interpret your data correctly. A model that scores well on AI Readiness isn't just well-documented - it's a model that AI tools can work with confidently, producing more accurate results for your users.

If you're building ontologies for MCP server support, Model Health is where you validate that the underlying models are ready for that level of AI integration.

What It Does

Model Health analyzes your model and produces:

Score cards across the top - AI Readiness, Model Quality, Documentation, Relationships, Structure, and Ontologies - each summarizing a different dimension of model quality.
Tables tab - a table-by-table breakdown of column counts, measures, missing descriptions, AI readiness, and quality scores.
Findings tab - a prioritized list of issues grouped by severity (Error, Warning, Suggestion), each with a code, explanation, and count of affected objects.
Relationships tab - an overview of all relationships for auditing cardinality, filter direction, and active/inactive status.

Two Types of Analysis

Run Analysis performs a rules-based scan checking for common structural and formatting issues like missing date tables, default summarization, bidirectional relationships, and orphaned tables. Runs locally with no AI key required.
Run AI Analysis adds a deeper, AI-powered review that catches issues pattern-based rules can't - like suspicious descriptions, inconsistent data types across related tables, or commented-out DAX. Requires an AI key configured in Settings > AI or through BYOK.

Scoring

The AI Readiness and Model Quality scores are each a weighted sum of individual factors (0–100 points). Click the (i) icon on either score card to see the exact breakdown and how each factor contributes. Weights are configurable in Settings > Modeling.

Source Selection

Use the Source dropdown to analyze your local working model or a model published to one of your configured environments (e.g., Develop, UAT, Production).

User Guide

For a walkthrough of running an analysis, interpreting scores, and working with findings, see Using Model Health.