Semantic models need a usage check before cleanup
Reviewed by Helen Jones
Chris Webb demonstrated the Semantic Link Labs find_unused_objects function for Power BI semantic models on 20 September. It can assess whether tables, columns and measures appear unused by examining downstream-report structure or DAX queries captured through Workspace Monitoring; the result can also be returned as a dataframe from a Fabric notebook.
Semantic-model cleanup often starts with a sensible aim—remove dead columns and measures so reports are easier to maintain—but a zero-use result is not a deletion decision. A field can be absent from the reports you inspect yet still be used by a separate query, a less-frequent process, or a future report change. The practical value is a repeatable review queue rather than an automatic cleanup tool. Combining report-structure evidence with observed query activity gives a reporting owner something to check with the business before removing an object, and lets them state the period and coverage their decision relied on.
Analysis
In a development workspace, choose one hard-to-maintain semantic model and run a usage review before deleting anything. Record which evidence method and monitoring window you used, then have the model owner validate the shortlist.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "Semantic models need a usage check before cleanup", Collab365 Spaces.