Every unused column and orphaned measure makes a model bigger, slower and harder to trust. Here is how to find them and decide what to cut.
Models grow by addition. A column gets added for a report that never ships, a measure is copied to try something and left behind, and six months later the model is twice the size it needs to be. Nobody decided that. It just accumulated.
The cost is real on three fronts. The file is bigger and slower to refresh. The field list becomes a thicket, so the next person builds the wrong thing because they cannot tell which of four similar measures is the right one. And trust drops, because a model full of things nobody can explain is a model nobody quite believes.
Find what is actually unused
The lineage and model views in Power BI show you what connects to what, which exposes the orphans. For anything serious, the Best Practice Analyzer in Tabular Editor flags unused columns and measures against a set of rules, and external tools such as MeasureKiller trace which measures and columns are used by no visual anywhere in the report. Between them you get a list rather than a hunch.
Decide before you delete
Not everything unused is safe to remove. A measure may be called by another measure, or a column may feed a relationship you cannot see at a glance. Check dependencies first, remove in a copy, and confirm the report still behaves before you overwrite the original. The goal is a model where every field earns its place, not a smaller model that is subtly broken.
Make it a habit, not a rescue
A model tidied once drifts straight back. The teams who stay clean run a quick review at the end of each build: anything added and not used comes out before it is published. It takes minutes when it is routine and days when it is a rescue, and it is the quiet difference between a model that lasts and one that gets rebuilt in a year.
Related reading
Power BI reports can now write back: what that actually changes
Row-level security done right: when everyone should see only their own numbers
How do you make Azure data pipelines reliable? Observability, monitoring and the checks that matter
Maisie Hudson
Data Consultant
Part of the Hopton Analytics team, delivering governed analytics programmes for UK mid-market organisations.
