Hi,With a large dataset, I encountered some relationship issues. Indeed, some tables do not filter the main fact table: project. The main goal of my dashboard is to show project information. The user should be able to filter these projects by using filters on the DIM table. Let's take a look at a sample that corresponds to my main issue:- Project is my main table I want to use in visuals.- Product is mainly a table used in slicers.- To avoid a many-to-many relationship between these two tables, I have a bridge table between product and project. However, as the directions are Product -> BRD Product and Project -> BRD Product, the filters from Product do not propagate to Project.Sample link: https://drive.google.com/file/d/1TsJ6mRvFgFoaT7u8aot3bihLP9q4JF4V/view?usp=sharing I thought of three solutions:- The quickest method could be to make Project->BRD product a double-sided relationship. However, if I do that in my large dataset, I could lose performance. Right?- I could force a specific relation with a DAX crossfilter; however, this would mean that I must create a measure for every Project attribute I want to use to change the direction with BRD Product. I do not think this is the most valuable solution.- The other solution could be to flip the direction to have BRD Product->Project instead of Project->BRD Product. However, Power BI does not allow me to do that. The tables get back to their original direction when I make a change. Both "update or delete relationship while actualizing data" and "Detect automatically new relations when data are loaded" can be checked or unchecked; this does not change the problem. Do you have an hint?Thank you !

Hi TheoAu, I understand your situation. Since the client requires slicers to drive the dashboard, enabling bi-directional filtering on the bridge relationship is the most practical way to achieve this, even though it isn’t ideal for performance or model simplicity. A good compromise is to apply it selectively only on the specific relationships needed for the slicers rather than across the entire model. This way you get the required functionality while minimizing potential risks with ambiguity or slowdowns. Thank you.

Data modeling: flip relationship direction | Microsoft Fabric Community

Forum Discussion

Related content

Direct Lake Model - Calculated Column/Tables

Direct Lake Semantic models Deployment

semantic Model direct lake

Warehouse security when using Direct Lake models

Relationship