Is Kimball dimensional modeling still relevant in a modern data warehouse?

So in Kimball modeling there were some assumptions made in1996 that just aren’t true in 2019:

  1. Databases are slow and expensive
  2. SQL language is limited
  3. You can never join fact tables because of one to many or many to one joins
  4. Businesses are slow to change

A lot has happened:

  1. databases are now fast and cheap
  2. SQL has evolved to include many more date functions and window functions that cut down on ETL work required
  3. Looker solved this with Symmetric Aggregates
  4. Business doesn’t wait for you to finish your perfect DWH

So I respect the work of Kimball and there are some great tips and techniques in the DWH Toolkit that I use everyday. But I see too many “over engineered” data warehouses that don’t need to be that complex and don’t encompass new things happening in the business.

5 Likes