Data warehouse architecture is key to performance management

Business performance management (BPM) programs are only as good as the data feeding them. That's why industry experts stress the need for great care when designing architectures for data warehouses, which are the primary information repositories for BPM initiatives.

"The most important thing to remember is that a data warehouse architecture must be flexible enough to accommodate future needs," says J. Paul Kirby, research director with Boston-based AMR Research. "Once you install a data warehouse, users will enhance their requests for information until that warehouse breaks down."

Kirby also recommends building enterprisewide warehouses rather than relying on smaller data marts that may only serve individual departments. He notes, however, that the enterprise-level warehouse does not—and in fact probably should not—be built all at once. The best approach, Kirby advises, is to look for an area of the business with a glaring need for a data warehouse and build one to accommodate that, while ensuring it can be expanded to meet other needs.

"Performance management initiatives typically come from the executive suite wanting to get a snapshot of the business through a series of key performance indicators," says Darren Cunningham, director of data integration with Business Objects, a supplier of business intelligence and BPM software. "So we look at projects starting with a few users and then spreading outward, with those initial users determining which data goes into the repository."

As the number of users expands, so will the type of data that needs to be in the repository. To avoid future problems, companies must take steps to preserve data quality. Kirby says the first step is having users review the data before it goes into the warehouse to ensure everything going into it actually needs to be there. There also must be rules to ensure data that will be used by more than one set of users is entered in the same format and has the same meaning.

Business intelligence vendors—including the aforementioned Business Objects, as well as SAS and Cognos—have products called ETL (for extract, transform, and load) programs that can put data taken from source programs into the proper format before it goes to a data warehouse.

Bill Prentice, a technology strategist with SAS, says users should know that an enterprise warehouse doesn't have to be in a single location. It can consist of a series of smaller warehouses that are linked through a technology infrastructure. AMR's Kirby agrees that approach provides the flexibility that should be inherent to any data warehousing architecture.