Here’s a list of things to avoid when developing a data warehouse. These are not necessarily in priority order. The priorities depend on your project.
1. Avoid technology for technology’s sake. Focus on business requirements and goals.
2. Avoid not securing a high-ranking business sponsor. If you lose your current sponsor, immediately find a replacement in the business sector of your organization.
3. Avoid trying to implement the entire data warehouse all at once. Approach the data warehouse project as a series of integrated sub-projects (data marts), and deliver each sub-project as it’s completed.
4. Avoid expending excess energy and budget on structuring the data in the warehouse. Do not over-normalize (starflake schema). Your focus should be the best query performance that you can deliver and delivering a quality, easy-to-use set of user interfaces.
5. Avoid complicating the user view of the data; keep it simple.
6. Avoid lack of integration among the sub-projects (data marts). Use conformed dimensions, as explained in “The Data Warehouse Bus Architecture,” November 2007, InstantDoc 96926; do NOT build data mart silos.
7. Avoid including only summarized data in the data warehouse. Many data warehouse analysts will want to drill down to the detail data after viewing the summaries.
8. Avoid the misconception that business requirements are static. Become a part of the business user community and stay active, that way you’ll be able keep your finger on the pulse of the company and detect directional change, which you can then confirm and translate into modifications to the data warehouse.
9. Avoid the idea that the data warehouse, once established, can exist without user acceptance. If the data warehouse becomes an expensive albatross with no way to cost-justify, it will be mothballed and you’ll be out of a job!
