Forum Discussion
Best Practice for creating Static Dimensions Table?
I'm working on Dataflows and have a question about creating dimension tables for a fact table with 3 years of event history.
Context:
- Fact table with events from the past three years.
- Around 15 dimension tables to map to this fact table.
- Plan to use dimensional keys in the fact table to join with the primary keys of dimension tables.
My Question: For dimensions that are static over time, like Employee Type (Permanent, Temporary, Contract) and Employee Status (Active, Terminated, On Leave):
- Is it better to manually enter this data using the "Enter Data" feature in Power BI, or
- Should I create and import this data from a CSV file? We will pick these CSV files from SharePoint
- Should I bring fact table then keep dimensional columns and remove duplicates before adding index to get keys
- Is there an alternate approach like using DAX (SUMMARIZE, VALUES etc.)
I'm looking for the best approach considering performance and maintainability.
Thank you!
Hi GulianiG
Both approaches are ok in a dataflow, however in a dataflow you can't relate tables together nor you can use DAX.
I would go with creating dimension tables in Sharepoint hence when someone other than the owner of the Dataflow wants to make changes, he will do it in Sharepoint(only the owner of the dataflow can make changes on it).
Performance wise, Dataflows are sources of truth for semantic models built on top of them, so they should be ok. Rather watch out the performance of the semantic models.
3 Replies
Hi GulianiG
Both approaches are ok in a dataflow, however in a dataflow you can't relate tables together nor you can use DAX.
I would go with creating dimension tables in Sharepoint hence when someone other than the owner of the Dataflow wants to make changes, he will do it in Sharepoint(only the owner of the dataflow can make changes on it).
Performance wise, Dataflows are sources of truth for semantic models built on top of them, so they should be ok. Rather watch out the performance of the semantic models.
You are welcome
