1. Educate yourself. I recommend getting Business Intelligence Roadmap by Moss, Atre and Youdon, and reading it cover to cover before you start.
2. Determine business requirements. Not all data warehouses are the same. You need to understand why the requestor needs a data warehouse. What are they trying to accomplish – saving time in collecting data, higher quality of data, supporting certain applications, making data available and/or enabling new company functions? You'll need to tie those objectives to what data sources you'll be pulling data from, what subject areas, what business rules you'll follow and what users and/or applications you'll support.
3. Make a timeline. Break up those business objectives mentioned above into 3-month incremental deliverables.
4. Choose architecture, methodology, and technology and compile a staff (and/or consultancy). You'll need to meet you business objectives as well as build your structures scalable so that you can add-on easily to meet future requirements. This includes securing business involvement with governance and stewardship programs.
More information on data warehouse development
This was first published in December 2007