Insurers Are Striking Gold With Data Mining Technology
“There is no doubt in our minds that, if done right, data mining will become a standard part of the business for any company,” says Rama Prasad, vice president of solution design for Chicago-based CNA Insurance.
At the same time, he notes that if a companys business process is not sound and fails to address data-quality issues, no data mining product will help. “As they say, junk in, junk out,” he says.
Prasads remarks are in accord with the views of several other executives at large insurance companies that engage in data mining.
Data mining involves sifting through a companys business and customer data to uncover patterns and relationships that could be useful to the business. Data mining can be done manually, or via computer programs that analyze the data automatically.
Vincent Armentano, president of workers compensation claims for Travelers Insurance, Hartford, Conn., calls data mining a “core strategy,” at least for his division. He says that data mining allows his division to go back into its data and to “use that hindsight to give us some foresight into what that next claims might look like.”
Armentano adds that data mining “brings you to the right event, but then what you do” with that event or how you handle it is up to you.
Prasad says that data mining “helps you get to the truth.” He explains that, while an insurance company may have based some decisions on “anecdotal evidence or hearsay,” data mining now can provide “data that you can trust.” As a result, a companys “decision-making becomes much more objective and very qualitative,” he says.
Richard L. Van Schoick, vice president of global marketing research for Prudential, based in Newark, N.J., says that data mining has been a “financially worthwhile pursuit” for the company.
He explains that Prudential can concentrate limited resources on maximizing its results. This means, for example, that Prudential “can spend less and offer more product to people more likely to buy.”
Gary Stromberger, director of decision support services for State Farm, Bloomington, Ill., says his company is “still in the learning stage” regarding the potential of data mining.
As a large company with large volumes of historical data, he says, State Farm is “very interested in the type of patterns, the types of observations that can appear to us by working and analyzing the data.”
Stromberger adds that the companys focus has actually been on data improvement initiatives to beef up its analytic capabilities.
Larry Koenen, a technical infrastructure specialist responsible for State Farms overall decision-support strategy, says the company has used data mining “as a tool within an overall information-understanding or analytic perspective.”
While not viewing it as “the end all,” State Farm considers data mining “a valuable way to sift through large amounts of information” that can then be validated through other means, Koenen says.
Bahman Dehkordi, also a State Farm technical infrastructure specialist, adds that the company does not rely on data mining alone for its decision support or decision-making. “Its generally a combination of tools and technology that we use,” he says.
Stromberger believes data mining “produces thought starters–something you never thought of before–and gives you the ability to investigate things further.” In short, “data mining gives you the capability of identifying questions you didnt know to ask,” he says.
Prasad says data mining has become “a standard part of[CNAs] architecture.” He reported that CNA uses data mining “broadly speaking torun our business.” More particularly, the company uses it for, among other things, “identifying all the key producers,” for marketing purposes and for customer service.
Data mining also helps CNA, which consists of multiple business units, to get “a very comprehensive and unified view” of its entire business, including its agents, according to Prasad.
One of the uses to which State Farm puts data mining is “market segmentation,” Stromberger says. “When weve introduced new programs, we can use that to look at their effectiveness.”
The global marketing department at Prudential primarily uses data mining “to identify key characteristics of products and customers, to indicate how they may react to things,” Stromberger states.
One example of this, he says, is that Prudential predicts the likelihood of customers switching carriers. The company then takes action to try to prevent that from happening.
Prudential now also predicts the likelihood of customers purchasing additional products. This helps the insurer to limit its “marketing dollars to those people most likely to respond to our offers,” Stromberger states.
The workers compensation claims division at Travelers uses data mining to spot so-called red flags, occurrences such as automobile accidents that are likely to be “subrogation candidates” or patterns of fraud, Armentano says.
Additionally, more advanced data mining tools have allowed Travelers to move “to more of a predictive mining process.” Armentano says that, for example, Travelers now tries to identify which injuries would benefit from nurse case management. In fact, Travelers is pursuing a patent for this, he says.
But, as Prasad notes, quality of data “is a problem that everyone faces in the industry.”
In fact, according to Van Schoick, historically in all institutions (financial, manufacturing, etc.), “the quality of data has always been questionable.”
Prasad stresses the importance of keeping data “clean,” noting that this is not a one-time project. “You need to have the right business processes to keep the data clean on a continuous basis,” he suggests.
Van Schoick says that a company must “use business logic and business knowledge in conjunction with what comes out of the data mining exercise.” A company must decide for itself whether the data turned up by data mining “makes business sense,” he says.
Similarly, Prasad cautions that it is necessary to “draw the line somewhere” when it comes to data quality. Data that is 95% correct may be “good enough” sometimes, while in other cases it may need to be closer to 100% correct. “You have to analyze each of those [situations] and decide where you want to spend your money,” he observes.
CNAs “ad hoc reports” are generated with a data mining tool known as Business Objects, produced by a company by the same name that has its North American headquarters in San Jose, Calif., Prasad explains. For its customized analyses, CNA uses Java programs to extract data from a data warehouse, he says.
Armentano indicates that for several years Prudential has been using data mining software called Enterprise Miner, from SAS Institute Inc., of Cary, N.C.
Stromberger declined to identify which software State Farm uses.
Travelers uses its own proprietary data mining software, although it keeps an eye on improvements in vendor products, Armentano says.
Prudentials Van Schoick notes that one of the problems with early data mining software was that users tried “to derive a lot of information from very little data.” Back then, data mining products could not handle “more than little bits of data,” he says.
But data mining products “have improved drastically over the past couple of years,” Van Schoick adds. “We can throw tons of data against these products and find idiosyncrasies [or]pieces of information imbedded in all the data that help out the business.”
He also notes that as data mining products improve, more data can be fed into them, and more information is obtained “out the back end,” making the products even more valuable to the user.
Travelers Armentano agrees. “As you get more sophisticated data mining tools being developed in the marketplace, as well as internally, you can get a better outcome, more refinement,” he says.
Van Schoick notes that a companys “own proprietary data is probably the richest source of information it can ever have, because its unique.”
Therefore, the more a company “can understand about how that data works with regard to the bottom line,” the better off the company will be, Van Schoick says.
E.E. Mazier is an assistant edtior for National Underwriters Property & Casualty/Risk & Benefits Management Edition.
Reproduced from National Underwriter Life & Health/Financial Services Edition, October 7, 2002. Copyright 2002 by The National Underwriter Company in the serial publication. All rights reserved.Copyright in this article as an independent work may be held by the author.