Knowledge is power, and whether you’re a department head in charge of a critical database or an admin guarding the boss’s cell number, the instinct is to cling to that power, not give it away.People often don’t even consider that your sales figures and the guy down the hall’s inventory figures and marketing’s campaign records could, if combined, provide information that is greater than the sum of its parts.
Like the blind men in the Indian legend, no one quite knows what the enterprise elephant really looks like, but everyone “knows” what the part they were touching suggests.
HP ran into that problem and it was costing it a lot of money, said Greg Battas, program architect and HP distinguished technologist, during a presentation at the SAS Institute’s BetterManagement LIVE conference.
A combination of what he described as the “tyranny of the urgent” and an extremely decentralized structure meant that, among its many divisions, it had more than 200 instances of SAP, 72 ERP installations, 740 business intelligence (BI) assets such as data warehouses and 30 enterprise data warehouses.
There were no common KPIs (key performance indicators) and governance focused on tactical needs, not strategic direction.
Then CEO Mark Hurd wooed CIO Randy Mott away from Dell, handing him a three-year, billion-dollar project to transform HP’s IT infrastructure.
Mott launched his enterprise data warehouse project in January, 2006, said Battas, and scheduled the first deliverables for April 30. New functions were to be added every three months after that – a major stretch. But, noted Battas, “having a new CIO while having to cut IT costs more than in half is really motivating.”
The entire elephant is now starting to take shape, and Battas offered key learning from the experience.
The first: centralize. Use one database, with one data model, and pull in data from all subject areas. Provide a normalized view, plug aggregate data for reporting, and only allow people access through a common presentation layer. HP is retiring its old BI systems as it rolls out the new one; people won’t have the opportunity to sneak back to their old ways. There have been some complaints from users who have lost functionality during the first phase of the rollout, but, said Battas, “there was a lot of consensus building, but at the end, there wasn’t a vote.”
Second, consolidate the best possible team from inside and outside the company to implement the project. At HP, Battas said, most of the managers building the warehouse came from outside the company, so it was built “the way it should be, not the way it’s always been.”
Third, make data suppliers responsible for data quality. “Getting the data was not hard,” Battas noted. “It was the quality of the data that was killing us.” Although everyone obeyed the letter of the law and provided data in the right format, a three-week delay ensued after CEO Hurd queried the new warehouse and got a gibberish answer. People were pulled from the source departments to clean up their correctly formatted mess.
Fourth, focus resources by stopping projects. When Mott arrived, there were 3,500 IT projects in the queue. He created a Top 10 project list, and pulled resources from other projects to accomplish those tasks. He also found “shadow IT” personnel, who worked in other departments performing IT tasks, and pulled them into IT.
Fifth, minimize dependence on outside organizations. “We have 80,000 people. If one of those 80,000 is not smart enough to do the job, we need to change people,” Battas said. HP used external products when necessary, however.
Finally, over-commit and under-deliver. “HP tends to over study,” Battas noted. “We were told early in the project, don’t let technology be an excuse for not being successful.”