Get certified for free when you join Fabric Data Days 2026 and dive into Fabric, Power BI, SQL, AI, and other essential data skills.
Join nowTry your skills in the Power BI Dataviz World Championship! Round one ends June 26. Join now
All - I have a test data model design here where I have multiple normalized dimension tables but the measures don't seem to give me the right counts. When you have many dimension tables like this that have many to many's, is this the way to properly setup the data model, or is it better to try to design a different, denormalized design?
Don't attempt to join fact tables. They should only be controlled by common dimensions. Don't have 1:1 relationships. Don't have bidirectional relationships unless it is absolutely unavoidable.
@lbendlin I got rid of the fact to fact joins. The 1:1 i will merge together. But is there any better option than still having this long string of bidirectionals to span all the factless fact and dimension tables?
As long as it remotely resembles a star schema you should be good. The data model needs to follow the business question, don't expect a single data model version to anwer all of yours.
Don't miss out on Data Days, June 15 through August 7. Learn Fabric, Power BI, SQL, AI and more.
Check out the May 2026 Power BI update to learn about new features.
| User | Count |
|---|---|
| 24 | |
| 22 | |
| 21 | |
| 20 | |
| 14 |
| User | Count |
|---|---|
| 58 | |
| 52 | |
| 41 | |
| 33 | |
| 31 |