Enterprise Company Dataflow Culture and Set-Up

[Kwarre35]

Hello, 

I am trying to find best practices on logically setting up Workspaces in terms of Dataflow Culture for an Enterprise Org. We currently have a Workspace = 'Master Data Dimensions' that we are using as a staging/process workspace to land all of our Re-usable Dim Data, but we are not quite sure how we want to organize/structure Fact Table Dataflows into Workspaces. 

Should we do this by Business Unit? (I.E. Finance, Supply Chain, Sales, Marketing, etc.)

Should we do this by Main Source System coming from Enterprise Data Warehouse?

Should we create seperate Workspaces for each Fact Table? 

Any thoughts/experiences or articles you can share would be much appreciated. 

Below is microsoft docs I have reviewed so far, not quite hitting the mark for me. 

https://learn.microsoft.com/en-us/power-bi/transform-model/dataflows/dataflows-best-practices

https://learn.microsoft.com/en-us/power-query/dataflows/best-practices-for-dimensional-model-using-d...

[Kwarre35]

@lbendlin  Good Questions.

Why we are using Dataflows to:

1. Reduce repetitive work

2. Enable Sepearation of Duties (MY department, IT, is responsible for Dataflows)

3. Reduce Workload on Source Systems

4. Consolidate refresh operations

5. Re-use Common Data

We are using Dataflows largely to deliver the data to the business analysts (dataset creaters) in a similar model as the data warehouse, to set up for Star-Schema Modeling success.

[lbendlin]

Does your company have any other base tool for data subject management? Where does your "common"  data reside?

Our situation is such that Power BI is only a small part of our data landscape, so "Re-use common data"  is a pipe dream for us. In addition we see dataflows mainly as a way to shield report developers from slow data sources. Any sufficiently performant data source does not warrant the additional steps in the process caused by the use of dataflows. 

[lbendlin]

Ask yourself

Why are you using dataflows?

What are you using them for?

Keep in mind that dataflows are basically bunches of Parquet files.