Data Marts, Fact Tables, Business Processes 1

[monsky]

Okay, this has been nagging for me for sometime even after reading through Kimball's book.

In Kimball's definition when you say data mart, it's simply a subset of the entire data warehouse. And a data mart is focused on a single business process. Does it follow that one data mart stand for a single fact table?

 

[johnherman]

Does it follow that one data mart stand for a single fact table?

Not at all. First of all, you could have aggregate or summary tables surrounding a single granular fact table. But even moreso, you could have multiple granular fact tables per Subject Area (datamart).

This is illustrated on page 94 of Kimball's "The Data Warehouse Toolkit Second Edition", but I'll recap here. The Procurement datamart could have fact tables for Purchase Requisition, Purchase Order, Shipping Notification and Tracking, Warehouse Receipts, and Vendor Payment.

I prefer to think of fact tables as representing business processes and datamarts containing subject areas. There could be multiple business processes within a subject area.

============

-------------------------
John Herman is available for short and long term data warehousing consulting and contracts.

 

[monsky]

Ok, thanks, that somehow declogs the arteries in my brain. Here's another one: can MOLAP cubes be considered data marts?

 

[johnherman]

It can help to consider them that way, but keep in mind the example with Procurement. The cube might not contain facts from all those business processes. So you might have more than one cube per data mart.

-------------------------
Sometimes the grass is greener on the other side because there is more manure there !