Dimensional Modeling: A Bottom-Up Approach

Presentation Summary

Neil Belcher; Cornell University

Dimensional Modeling: A Bottom-Up Approach

The Challenge: How do you design and implement a Dimensional Datamart in parallel with the ERP’s implementation? During any ERP implementation the majority of focus is on the application implementation and the BI solution is often an afterthought. How can you design a datamart without clear requirements from the functional leads? Can you get reliable answers to your functional questions when the users are learning the new system themselves? Do you have any sample data specific to your implementation to design from? Even without clear requirements, my experience has shown that the application data can lead you to a foundational datamart design based upon irrefutable facts and obvious dimensions in the application. I call this the “Bottom-Up” approach to data modeling. Traditionally, a “Top-Down” approach is used where clear requirements are first gathered and then synthesized into a conceptual, logical and physical design. The “Bottom-Up” approach uses the reverse order by first examining the physical source and using that to derive the physical/logical designs. Later in the project, a parallel “Top-Down” approach is conducted to refine what is delivered from the “Bottom-Up” approach.