Forum Discussion

MohamedFowzan1's avatar
MohamedFowzan1
Super User
1 month ago

Feedback for Semantic Model Health Dashboard

Looking for feedback from the Power BI community before I officially share this more broadly. I've been working on an Automated Semantic Model Health Platform built using Microsoft Fabric, Semantic ...
Need Help
Prince0011's avatar
Prince0011
Solution Sage
1 month ago

This is a great initiative and a very practical use case for combining Microsoft Fabric, Semantic Link (sempy), and semantic-link-labs for enterprise Power BI governance.

A few additional areas that could make the Semantic Model Health Platform even more valuable:

1. Expand health scoring beyond performance

In addition to BPA rules and optimization checks, consider adding governance metrics such as:

  • Unused or rarely used semantic models.

  • Dataset refresh failure frequency.

  • Refresh duration trends.

  • Model size growth over time.

  • Number of calculated columns vs measures.

  • Number of unused columns/tables.

  • Duplicate measures and inconsistent naming patterns.

  • DirectQuery vs Import model recommendations.

2. Add dependency and impact analysis

Since many organizations have hundreds of semantic models, it would be useful to include:

  • Report-to-semantic-model dependencies.

  • Downstream usage count.

  • Last accessed date.

  • Business owner/team ownership.

  • Impact analysis before making model changes.

3. Historical trend analysis

The historical health score tracking is a great feature. Some additional trends that could help administrators:

  • Health score improvement/regression over time.

  • Top recurring optimization issues.

  • Teams/workspaces with the highest technical debt.

  • Capacity impact before and after optimization.

4. Include adoption and governance signals

For enterprise environments, technical health alone may not be enough. Consider adding:

  • Workspace activity levels.

  • Report view counts.

  • Certified/promoted status.

  • Sensitivity labels.

  • Ownership and documentation completeness.

One more interesting direction could be using AI capabilities to generate optimization recommendations, for example:

"This semantic model has 15 unused columns, 8 high-cardinality fields, and 12 unused measures. Removing them could reduce model size by approximately X%."

Overall, this is a strong example of how Fabric can move Power BI administration from reactive troubleshooting to proactive governance.

Looking forward to seeing how this evolves. Great work!

💡 Helpful? Give a Kudos 👍 — keep the community growing.

✅ Solved your issue? Mark this as the Accepted Solution ✔️

Best regards, Prince Singh | Data Science & Microsoft Fabric Enthusiast