As insurers emerge from the distraction of getting their computer systems Year 2000 compliant, they've had limited success turning their attention back to data-mining initiatives, , said Tom Chesbrough, of Kapstone Systems Inc.

Quoting statistics from the technology-research firm the Gartner Group, Chesbrough noted that 85% of the first generation of data-warehousing efforts have either failed or haven't achieved their goals. Chesbrough, speaking at A.M. Best Co.'s "Fulfilling the e-Promise" conference in Boston Monday, said one reason is that many insurers have taken the bottom-up approach of creating "datamarts" instead of a top-down approach.

Building datamarts bottom-up means developing islands of databases that might be labeled sales/marketing, claims, finance and underwriting. This approach is small-scale and fast but easily results in "dirty datamarts," said Chesbrough, president and chief executive officer of Kapstone, a Kansas City, Mo.-based business intelligence company specializing in the insurance industry.

Companies need to produce centralized, enterprise-wide data repositories, but this process takes time. A narrowly focused datamart can be up and running in 90-180 days but uses limited data resources, often lacks standardization and relies heavily on summarized data, Chesbrough said.

A centralized datamart takes seven months to three years to do right, he said, because it uses many data sources, is scrubbed for consistency and is configured for easy access throughout an organization and remotely via the Internet, intranets and extranets.

Success depends on a company making a high-level commitment and installing a team leader with data-warehousing skills and experience, Chesbrough said. The benefits of success, however, can be enormous, he said.

"We're in a data-overload situation," Chesbrough said. "This is one of the major reasons businesses are turning to business intelligence," or data-mining.

For example, "many insurance companies don't really know how much it costs to sell a product," he said. But competition coming from all sides is forcing them to look at the billions of pieces of information they control to understand "the ultimate cost of trying to maintain their customer base in competitive times."

"Business intelligence," Chesbrough said, "is a process of assembling disparate data, transforming it to a consistent state for business-decision making, and empowering users by providing them with access in multiple views."