Data warehouses are still high-risk IT projects. In all, 41% of the companies surveyed by the Cutter Consortium, an IT consulting and market-analysis firm, have failed data warehouse projects, and only 15% call their data warehousing efforts to date a major success.

Data warehouse systems are complex and expensive; 60% to 90% of all data warehouse projects either fail to meet expectations or are abandoned. Project success rates have improved in recent years, thanks to maturing implementation methodologies and packaged data warehouse technology. This week, NCR Corp.'s Teradata division shipped a new release of its data warehouse software with tools that make it easier to build and manage large-scale data warehouses.

The Cutter findings, while an improvement, indicate that assembling a data warehouse remains a risky proposition. "They're still tricky to pull off," says Curt Hall, a Cutter senior consultant who oversaw the detailed survey of 142 companies conducted in 2002. Technology remains part of the problem: Only 27% of survey respondents were confident about the current state of data warehouse technology, while 58% were "cautious," Hall says. Only 23% use packaged data warehouse software; 77% developed their own.

Process may be as much of the problem as technology. Only 25% of respondents had a well-developed enterprise business-intelligence plan when they began their data warehouse projects, Hall says, and many do a poor job understanding the needs of those who ultimately use the system. And just 18% conduct return-on-investment studies once data warehouses are completed.

Despite the problems associated with data warehouses, 49% of the surveyed companies have at least one data warehouse in use and 70% are building one. Jenny Craig International will go live next month with a data warehouse to store 3 million client records collected from its weight-loss centers. The company's IT executives are confident the project will succeed because of the effort devoted to researching available software and interviewing the system's expected users about their needs. Says IT manager Sophia Ruiz, "We did as much as we could up front."