0

I'm new to dimensional modeling. I'm reading Kimball's "The Data Warehouse Toolkit".

As soon as I understood, Conformed Dimensions are a good thing, a key concept for integration of different fact tables. Usually you will have a separate fact table per business process, and if someone want to make decisions based on multiple processes - in most cases conformed dimensions allow to perform Drilling Across instead of Consolidated Fact Table. Looks pretty straighforward.

But how Abstract Generic Dimensions differs from Consolidated Dimensions?

Looks the same for me. For some reason Abstract Generic Dimensions are consider anti-pattern. The referenced example says it's bad to use the same geo-location dimension table for employees, customers and vendors. Two reasons: attributes may differ and dimension table size. But haven't Conformed Dimensions got the same downsides?

VB_
  • 45,112
  • 42
  • 145
  • 293

1 Answers1

2

The Kimball article on abstract generic dimensions doesn't say anything about dates.

A Customer dimension shared across sales and marketing facts is a conformed dimension

A single Person dimension containing employees and customers is an example of an abstract generic dimension (which might be "bad" if there is very little commonailty in attributes or processes)

However IMHO nothing in Kimball is a hard and fast rule - I see it as guidance, and the note on abstract generic dimensions to me is just a warning to do proper analysis before jumping into using one dimension to model two things that seem the same but probably aren't from a data detail perspective.

Nick.Mc
  • 18,304
  • 6
  • 61
  • 91
  • thank you for the answer! My bad, corrected to "the same geo-location dimension table for employees, customers and vendors". What makes "Person dimension containing employees and customers" a bad dimension? The fact that employees and customers have small intersection? Or a logical (business process) difference between two: customers and employees attributes may evolve independently in the future? – VB_ Dec 23 '19 at 11:58
  • did you mean by your example that BI will never run decisions based on employees and customers simultaneously? – VB_ Dec 23 '19 at 12:00
  • 1
    It depends on your BI. Do business analysts asks questions of this nature? Probably not. Do you have the same level of detail (attributes and processes) around both? probably not. – Nick.Mc Dec 23 '19 at 12:02
  • I've added some more to my answer – Nick.Mc Dec 23 '19 at 12:04