And how to find them automatically
Most Power BI models have health issues that are hard to spot manually. They don't break anything immediately, but they create confusion, make models fragile, and become expensive to fix later.
Here are the 8 most common ones we see across real-world models.
M queries that reference C:\Users\someone\Desktop\data.csv work fine on the author's machine and nowhere else. When the model gets shared or published to the service, the refresh breaks.
Fix: Move data to a shared location (SharePoint, database, data lake) and update the M query source.
Inactive relationships are dotted lines in the model view that most people forget about. They were usually created by accident or left over from a previous design. They add confusion for anyone reading the model, because it's unclear whether they're intentional or forgotten.
Fix: If you need the relationship, activate it. If you don't, delete it. Use USERELATIONSHIP() in DAX if you need it only in specific measures.
Power BI's time intelligence functions (TOTALYTD, SAMEPERIODLASTYEAR, etc.) rely on the model knowing which table is the date table. If no table is marked, these functions either fail silently or give wrong results.
Fix: In model view, right-click your date table and select "Mark as Date Table". Choose the date column.
Measures with magic numbers like IF([Amount] > 1000, "High", "Low") are a maintenance nightmare. When the threshold changes, someone has to find every measure that uses it. In large models, this gets missed.
Fix: Use a parameter table or a disconnected table for thresholds and reference that in DAX instead.
M steps like Table.SelectColumns(Source, {0, 1, 3}) reference columns by position instead of name. If the source schema changes (a column gets added or reordered), the query silently pulls the wrong data.
Fix: Always reference columns by name, not position. Use Table.SelectColumns(Source, {"Name", "Amount", "Date"}).
Tables that are hidden from report view and have no relationships to other tables. They're invisible to report authors, take up memory during refresh, and are usually leftovers from deleted features or experiments.
Fix: Check for hidden tables with no relationships. If nothing references them, delete them.
Power BI allows two measures with the same name in different tables. This creates ambiguous references in DAX and confuses report authors. [Total Sales] might resolve to either measure depending on context.
Fix: Rename one of the duplicates to make each measure name unique across the model.
Columns and measures without descriptions are undocumented technical debt. When someone new inherits the model, they have to reverse-engineer what [CM_YTD_vsBUD_pct] means from the DAX. This slows everyone down.
Fix: Add descriptions to your measures and key columns. Even a one-line description saves future you hours of guesswork.
You can check all 8 of these issues in seconds. Upload your .bim or .pbit file and get an instant audit report — plus full model documentation with lineage and impact analysis.
Your file is processed in memory and never stored. No account required to view the report.