Data Warehouse Or Bust

Owens Corning gave its IT group only 10 weeks to roll out an unconventional data warehouse and analytical package that would help rescue the company from a desperate gross margin situation.
Kalido's technology is based on a concept called generic modeling, which was a piece of "pure R&D" within Shell that Hayler ran across one day, and decided would be applicable to data warehousing. A Kalido data warehouse has "an unusual shape," according to Hayler. It has a single data template that holds all the data, as opposed to a traditional warehouse wherein users build schemas based on the queries every time they need data. In Kalido, the structure of data is actually entered into the warehouse as data. A Kalido system, therefore, is standardized and packaged, rather than custom-built. "Once you’ve defined your model," Mikkelson said, "you're done. Kalido will then describe all the structures and schemas automatically."

This also makes Kalido more flexible, said Mikkelson. If, for instance, a tax audit makes it necessary for a company to go back and alter historical data, those changes can be made without having to recreate the entire information world.

After funding was approved for the Kalido warehouse -- Mikkelson puts the dollar figure at "several million" -- the real challenge began. So dangerous were the company's gross-margin pressures that management wanted the warehouse installed almost immediately.

"They had an entertaining time schedule," says Hayler. Mikkelson and his team and a Kalido technician started the project on May 24 of last year. They were told that they had to be done by July 30. "So we had ten weeks to implement a fully functional infrastructure, a three-layer infrastructure, with the centerpiece being Kalido, and a gross margin analytical tool, at same time," Mikkelson said. "We did it in ten weeks, but anybody more sane than me probably wouldn't have taken it on."

In week one, Owens IS gathered requirements from end users. In week two, it began the design process, which it finished by the end of week four. The following week it installed the software, and over the next five weeks it went through five iterations. After each, IS got feedback from users and instituted those ideas. Meanwhile, in weeks seven through ten, IS began deploying the tools in the real world and training end users. The deadline was met.

Hayler said that the speed of the deployment resulted from Kalido's scalable structure. "Instead of asking questions and then going off to write code for six months, you can build the thing up gradually. You can get lots of feedback from users on a daily basis, and you can do a daily build of data."

In keeping with Mikkelson's seeming preference for obscure software vendors, Owens uses apps from Corporate Radar and Metreo to produce gross-margins reports and analytics. Mikkelson praised the small footprint and drill-down capabilities of Corporate Radar, a privately held vendor whose products are all Microsoft-based. Metreo, Palo Alto, specializes in product-pricing analytics.

The Kalido warehouse at Owens consolidates raw information from about a dozen ERP database systems, including SAP, Scala Business Solutions and J.D. Edwards, among others. Not including the IS department, there are about 30 or 40 Kalido power users at Owens Corning and about 600 users total.

The only metric Owens will use to gauge the success of Kalido is, of course, gross margins. Already, Mikkelson said, the software has covered its cost and then some: "The payback period has been less than a year."