Digging the Data Mine.

Data warehousing and modeling can be used as a small-commercial underwriting tool.

Warehouses traditionally have stored forgotten information. No longer. Today computers are mining warehouses for all they're worth.

The advent of data warehousing holds tremendous promise in improving loss ratios for the insurance industry. Used with traditional underwriting information and skills, data warehousing can help identify better risks in marginal classes and poor risks in good classes, enabling an insurer to make intelligent underwriting choices and price risks more accurately.

For nearly two years, Hartford Financial Services has used data warehousing to determine insurance score, which is used to help underwrite small-commercial business. Insurance score is the product of an objective, statistical analysis of business financial-scoring information. Much of this comes from a business credit-reporting agency, such as Experian or Dun & Bradstreet, and is combined with other data elements that identify the relative likelihood of an insurance loss. It does not measure creditworthiness or the ability to pay premiums.

Insurance score has been used to evaluate personal lines risks for years. Reviewing financial information has long helped underwriters of large-commercial accounts, but the small-commercial insurance side has been slower to adopt it. Used properly, insurance score enables underwriters of low-premium small-commercial risks to make more prudent, cost-effective underwriting decisions.

For example, a large-account underwriter reviews an application along with a financial statement that helps determine the financial soundness of the company Similarly, insurers that maintain a data warehouse can quickly access equivalent information for small companies and use it to make sound underwriting decisions.

Data modeling is just one tool to get the most out of the information in a data warehouse. In this case, an insurer uses the data to consider possible predictive characteristics, then compares them against actual claim performance from company records. The insurer notes and retains the characteristics that are most relevant and discards the rest. These characteristics are assigned weights and further tested against the actual results in the company's claim database. Eventually, the most predictive characteristics are prioritized and used to sort risks into categories by likelihood of loss.

Data warehousing and modeling can be used as a small-commercial underwriting tool to:

* help determine whether the insurer will accept or reject a particular risk, something permitted by all states;

* facilitate accurate pricing for an account in the all but a few states that permit this use. The underwriter still might decide to accept a marginal risk but specify that it be priced according to published rates;

* help validate an exposure base, such as type of operation, years in business or size. Any deviation from furnished information would raise a red flag and may cause the underwriter to look deeper into the application. A company might, for example, claim to be a painting operation in its application but could advertise itself as a construction company; and

* help agents seek out business likely to be acceptable to a particular insurer. Eventually, agents will have access to some of this information, saving them time and money. This also will allow agents to pre-underwrite, even before they make contact with a potential account.

In Hartford's case, an overwhelming majority of its accounts fall into acceptable insurance score categories. Data warehousing enables us to identify and avoid the 2% to 5% of cases that look good using traditional underwriting tools but are rated "poor" when put through the insurance score filter. By avoiding risks with much higher-than-average loss potential, we expect to save money and pass it on in the form of continued competitive rates and pricing. That's what enables an insurer to remain a strong, healthy competitive market for small-commercial business.

Growth is another important benefit from data warehousing and modeling. It can help an insurer dispose of preconceived notions that are no longer valid predictors of risk. It can permit a company to expand its underwriting guidelines and grow.

Hartford has used this to help expand the number of classes eligible for our small-business package policy. In some cases, it enables the insurer to entertain marginal risks with a good insurance score. And eventually it will enable agents to focus on the best leads and increase their sales on behalf of the forward-looking insurers who use this technology.

Data warehousing is very much in its infancy as a commercial insurance underwriting tool; Hartford's experience has shown it has tremendous potential to help refine decision-making and expand our market potential.

Judith A. Blades is president of business insurance at The Hartford Financial Services Group, Hartford, Conn.