Using Datavault modelling before the use of Kimball methodology for a datawarehouse

Question

1. In a data ware house project why do we need to use DataVault modelling before transfer the data in data marts(which use Kimball methodology) from landing / staging area databases?. ie Why can't we use the Kimball methodology to transfer the data straight forward from the landing / staging area databases to final data marts?

2. Is it a must to do this?.

Answer 1

Well, you don't need to use the Data Vault methodology. Also, you don't need to use Kimball. Also, you don't need to use 3nf.

This all depends on the requirements you have in your environment. On

• the data structure,
• data complexity,
• sizing,
• schedules,
• changing source formats,
• need of historization,
• requirements for reports, dashboards or other ETL structures...

There is no 'need' to do Data Vault specifically.

It all depends on what you want to do and what your requirements are.

Answer 2

I've noticed that lately, the Data Vault 2.0 methodology has become quite prevalent in EDW projects. Kimball/Star-schema data models are still very much used, but they are mainly just the top level abstraction on top of the Data Vault that allows for reporting.

Kent Graziano has a great primer to understand DV modelling here, which is an excerpt from Dan Linstedt's book (also a great read).