Risk managers always have known crunching numbers can help them spot business problems, but not everyone realizes it is now much easier to do.

Technological advancements are making it possible for risk managers to analyze data culled from numerous sources -- inside and outside their companies.

Extracting meaningful information from numerous databases and cross-referencing it to find patterns, trends and correlations that might otherwise be overlooked is called "data mining." Assembling the information in one place is called "data warehousing." The objective is to analyze the information to make informed business decisions.

By definition, data mining usually focuses on legacy, or historic, data that is mostly transactional and financial, such as claims figures and reserves established.

"Opportunities are growing for data mining," said Chad Karls, an associate actuary in the Milwaukee office consultants of Milliman & Robertson Inc.

Unfortunately, few risk managers seem to be aware of data mining opportunities, other experts point out.

"If you asked risk managers what their investment is in data mining and data warehousing, they'd say not much," said Mark Dorn, president of DORN Technology Group Inc. in Livonia, Mich.

Yet many risk management solutions already are "holistic," derived by assembling information from numerous internal and external sources, he pointed out.

For example, employers can eliminate redundant disability expenses by cross-referencing health and workers compensation claims data.

The MEDSTAT Group helped one employer client do this shortly after the employer started cracking down on short-term disability claims.

"We found a lot of their workers did have claims in both (the STD and workers compensation) programs at the same time, and the ones that were overlap claims had a higher incidence of attorney involvement and much higher cost," said I. Jeff Turshen, consulting manager in Stamford, Conn.

MEDSTAT is conducting similar studies for other employer clients on the impact health promotion programs are having not only on health claims, but workers compensation claims, absence and turnover rates as well, according to Mr. Turshen.

The sources of information risk managers and their consultants can tap for data mining purposes are increasing constantly with the growth of the Internet and personal computing technology, experts say.

Information can be downloaded from the Internet or transferred directly via phone line from an external source, such as an insurance company database. Numerous database files also are available on disk, tape, cartridge and CD-ROM from government and private industry sources.

For example, "OSHA is a good source of data on injuries and incidence rates," Mr. Dorn said.

Other sources of pertinent external information available in a variety of formats include: third-party claims administrators, insurers, the Bureau of Labor Statistics, the National Council on Compensation Insurance, the American Compensation Assn. and the National Committee for Quality Assurance.

"A lot more data is readily available," observed Tim Beck, a principal at Buck Consultants Inc. in Los Angeles who specializes in health risk management.

Rick Betterley, managing director of Betterley-Donoghue, a Boston-based risk management consulting firm, said he recently conducted an employment practices liability exposure study for a client using information obtained from the Internet.

"The information was literally sitting there so that anyone could download it with a few mouse clicks," he said.

But Mr. Karls of Milliman & Robertson recommends that risk managers who pull information from the Internet take the time to verify its authenticity.

"Internet information integrity is in question," he said.

Still, he acknowledges there are benefits to using Internet data: It will be updated throughout the year, while information from government databases or other statistical gathering organizations such as A.M. Best Co. usually is updated annually.

Internally, risk managers usually have access to data on claims, payments, reserves, exposures, litigation and incidents.

While this information sometimes is assembled on a risk management information system, it often is being tracked separately.

"In a typical risk management office, there's a lot of information. But it's not always in one place," Mr. Dorn said.

Most data mining projects can be easily completed by inputting information from multiple databases into a single database program, such as Microsoft Access. The data miner then can test various assumptions by sorting the data in various ways in a spreadsheet program.

It's become easier for users to assemble information from numerous databases because most new database software programs adhere to the Open Database Connectivity Standard, according to Mr. Dorn.

Some such easily transferable "data fields" -- which Mr. Dorn defined as types of information -- are dates, ages, birthdates, ZIP codes, and financial information, such as payroll, sales and claims data.

By assembling the numerous data fields into a single database, risk managers can spot hidden patterns, trends and correlations.

"You also do come up with some surprises," said Mr. Turshen, who recently worked on a project involving Kentucky Fried Chicken's workers compensation claims experience.

The project, which required assembling data from human resources as well as risk management and other corporate sources, was highlighted at the recent Business Insurance Workers Compensation Conference in Santa Monica (BI, Nov. 10).

"One of the surprises at KFC was the inner-city stores had a fairly good track record, and then, as we dug deeper, we found that the reason was that the managers there tend to stay longer," Mr. Turshen recalled. "So you do have the impact of multiple variables on each other."

Another MEDSTAT study produced similarly surprising results, according to Patricia Insley, marketing manager in Ann Arbor, Mich.

In studying the relationship between supervisor feedback and employee absence rates, researchers found that greater feedback actually resulted in higher absence rates, not lower, as was assumed.

The findings caused the researchers to question the data, because "this was very counter-intuitive," she said.

But, after further examination, "what we found out was that the kinds of feedback that supervisors gave was productivity-based. . .so people saw this as a very painful experience rather than as a very encouraging experience," Ms. Insley said.

Risk managers should establish specific objectives before assembling a data warehouse from which to mine information, experts advise.

"If you ask risk managers how much data they want moved, they'll all say, 'All of it,' " Mr. Dorn quipped. But all of it may not be pertinent, he said.

"They need a plan."

Mr. Beck concurred.

"You need the right set of assumptions, or you'll end up with a flawed result," he said.

Often the data mining objective will depend on a risk manager's background.

For example, a risk manager with a claims background may want to drill for information that will help him or her set more precise reserves, Mr. Dorn suggested.

But a risk manager with a more business-oriented background may want to use the data to forecast future claims frequency and/or severity using numerous growth scenarios.

Other risk managers may want to use the data to do benchmarking, comparing their companies' experience with industry performance standards.

Perhaps a risk manager will want to use the data to conduct a special study, such as the frequency of back-injury claims among workers over 40.

"You need to be clear which of these areas you want to focus on before starting," Mr. Dorn said.