Best Practise question for Modelling in Powerbi/Tabular - Avoiding Chained/Nested Chasm Traps

[Anonymous]

I am currently trying to understand the best practises around Powerbi and Modeling in Powerbi/Tabular (With big data). Below is an example of a snowflake design. (Client  which filters matter which then filters the facts). I just wanted to make sure i am following best practises. 

My approach to a re-design is to split Client off Matter and have it directly pointing to the fact which will create a star schema Model. Is this approach correct? in saying that, the result of me creating a star schema will cause the Client to stop filtering items on Matter unless i created Bidirectional filtering from the fact to both client and matter. What would your approach be? I do not know the implications of having a huge model with bidirectional filtering which supports the star schema design.

[Anonymous]

Another recommended approach is to merge your Client and Matter table into one single denormalised table.  Since both of your tables need to link via the Matter ID, this relationship will still work and you can still organise filters by the Client data.

[Anonymous]

I was considering that idea, however my matter and client dimensions are pretty big. So for end users i think i need to keep them split up. 

[Anonymous]

Being big is actually the reason to merge them.  The way the engine stores and compresses the information will likely mean it won't make a large change to the model size, but there is likely to be a performance gain.

From an optimisation standpoint, you can remove unnecessary columns, although from your screenshot you seem fairly lean already, which is a good thing.

[v-gizhi-msft]

Hi,

Please try to make 'Dim Matter' to 'Fact Debt' and 'Fact Sales' reltionships to inactive and create relationships between 'Dim Client' and other two fact tables:

Hope this helps.

Best Regards,

Giotto Zhi