Choosing tools for model quality and performance
1. Choosing tools for model quality and performance
In the last lesson, you kept the model's data current. A current model can still have design flaws or feed slow reports. In this lesson, you'll meet three tools that catch these problems, and when to use each.2. Three tools, three stages
Model quality and report speed are checked at different stages, and each stage has its own tool. While you build a semantic model and decide how its tables relate, Best Practice Analyzer checks the design against modeling rules. Once a report is complete, Performance Analyzer measures how long each visual takes. If a slow visual traces back to a measure, DAX Studio helps you dig deeper into its query.3. Best Practice Analyzer checks model design
Best Practice Analyzer belongs at the start, while you're still shaping the model and deciding how tables relate. It checks the model against design rules, such as whether numeric ID columns are summarized by default, which can produce misleading totals. In Fabric, open the semantic model, and in the Model health group on the Home tab, select Best practice analyzer. Fabric creates a notebook with the analysis ready to run. When it finishes, results appear in tabs such as Formatting, Maintenance, and Performance. Each row names a rule, the object it applies to, and a severity. Warnings mark issues worth fixing first, and information icons mark lower-priority suggestions. The same rules also run in tools such as Tabular Editor.4. Performance Analyzer: QA for finished reports
Once you have a semantic model, you build reports on it in Power BI. After a report is built, and before you publish it, Performance Analyzer records how long each visual takes, so you can find improvements before users notice. You won't build a report in this course, so you won't open Performance Analyzer here. In practice, it's under Optimize in Power BI Desktop, or under View in web Edit mode. Here, Top Customers takes 780 milliseconds, more than twice as long as any other visual, so it's the one to expand first.5. Timing categories show where to look next
The breakdown follows the work a visual does. First, the semantic model processes the visual's query, measured as DAX query. If the table uses DirectQuery, the model also waits on the external source, shown as Direct query. Power BI then draws the results, recorded as Visual display. Other covers the rest, such as waiting for other visuals. The largest category tells you where to look next: the model, the source, or the visual.6. Follow the largest timing category
Expanding Top Customers shows that DAX query accounts for 420 of its 780 milliseconds, so the next step is to review the measures behind the visual. That's where DAX Studio comes in. It's a free external tool: paste the visual's query, copied from Performance Analyzer, and Server Timings show where the time goes. If another category took longest, you'd look elsewhere: the external source for Direct query, or the visual's design for Visual display.7. Report design can add load
Report design can add load too. Every visual sends its own query, so a page with many visuals does more work each time it loads. A column's cardinality, its number of distinct values, matters as well. A slicer on CustomerID, with 1.2 million distinct values, has to query a very long list. Complex measures add query work too. Treat these as warning signs, and confirm them with Performance Analyzer timings before redesigning anything.8. Let's practice!
Now let's put these tools into practice!Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.