Twelve Steps to Revitalize a Data Warehouse
Tales from the front lines: How a BI director revitalized a failing data warehouse
Many data warehouses are launched with much fanfare and promise but quickly fail to live up to expectations. According to TDWI research, most companies do not abandon these less-than-successful implementations, but rather restart them using a new strategy, new staff, or new focus.
The following article contains verbatim excerpts from a presentation delivered by a BI director at the recent TDWI conference in San Diego. He described in vivid detail the steps he and his team took to revitalize a data warehouse that had failed to deliver significant business value, and had earned an undesirable reputation among many sectors of the business. Over a four-year period, the team transformed a moribund data warehousing program into a strategic initiative that has generated tens of millions of dollars in cost-savings and revenue gains.
1. Recognize you have a problem. “Our data warehouse was a dumping ground for all data from a few sources. It was too slow, too costly, had too much data and not the right data, queries took forever, and managers hated it.”
2. Get adequate investment to do a makeover. “My boss gave me approval to fix the data warehouse but said, ‘Don’t come back and tell me you need $6 million and three years to deliver it.’ So, I came back a few months later and said I need $18 million and six years. Anything less and we would have starved the initiative and wouldn’t have succeeded. He saw the rationale of the case I made and agreed to support me.”
3. Start Selling. “I got my tap dance shoes on and ran around to every vice president in the business – tax, audit, finance, etc. – and asked about their pains, problems, and needs. Some said ‘Thanks for asking.’ Others said, ‘Get out of my office! We should outsource the data warehouse!’”
4. Leverage Sarbanes-Oxley. “I rode SOX as much as I could. I told the VPs, ‘We need a single version of truth so we can the report data one way to Wall Street. You don’t want to go to jail, do you?’”
5. Be Memorable. “If you don’t stay in their face, they’ll forget about you. Set up a governance committee and invite someone in each organization who knows the data to sit on it.”
6. Turn Potential Enemies into Allies. “Tell committee members that they are geniuses and you want to tap into their expertise and that you are not interested in taking their data. If you can get them excited, they will spread their excitement to the rest of the organization.”
7. Over-Deliver Benefits. “Since you won’t get all groups to sign on right away, work with those who have the most interest. Make sure you deliver an order of magnitude more benefits than you promise. This gets the ball rolling.”
8. Accelerate the Snowball. “Go to the VP of the next group and say, ‘The VP I’m working with says that you should look at this solution since it might benefit you.’ Once the VPs see a success, they don’t want to be left behind. It’s kind of an ego thing.”
9. Plan for Growth. “You need a serious plan for scalability because if your system crashes or bogs down, you’ll immediately lose credibility with the business.”
10. Deliver the Data. “Initially, we created data marts for each group so they got just the data they wanted and query response times were fast.”
11. Consolidate. “Once our data volumes got large and the number of data marts became unwieldy, we consolidated everything into a single dimensional model that becomes an enterprise transformation layer that we used to populate all data marts and applications.”
12. Standardize. “It can be painful to get VPs to sign off on a consistent set of rules for enterprise data– it took us several multi-day offsite meetings—but now it’s stamped in blood. If the data doesn’t come out of the data warehouse, it’s not official. However, we also ensure that they can get localized views of that data."
About the Author
Wayne Eckerson is director of research at The Data Warehousing Institute (TDWI), a provider of in-depth education and research in the business intelligence and data warehousing industry.