Data flow: inexpensive third-party data and mathematical models now allow insurers to underwrite small and midsize commercial accounts as thoroughly as big risks and personal lines.

Underwriting insurance for small and midsize businesses has been something of a stepchild for the industry. Premiums per risk are typically small, so insurers can't afford to conduct the same level of intensive underwriting scrutiny as is typically done for large accounts. Nor is there the high volume that has allowed insurers to bring technology to bear in personal lines underwriting.

But small to midsize commercial lines underwriting is finally catching up with personal and large commercial lines underwriting through smarter use of technology. It's now feasible for underwriters to bring in third-party data automatically. With richer data, they can make better, more accurate underwriting and pricing decisions. This third-party data falls into the following three categories.

Validation. This includes data that validates information supplied by the insured, such as building valuations or the gross weight and original cost of a vehicle. Under the traditional underwriting model, the insurer simply relies on the application. But this information isn't always accurate, and it should be validated. For example, in commercial auto, rates are based on the original cost of the vehicle and its gross weight, which the applicant may not really know. If they're understated, the insurer won't collect an adequate premium. With access to a third-party database, the underwriter can crosscheck the application to make sure that the weight and cost are consistent with the manufacturer's numbers. Similarly, outside data can be used to validate the value of a building. Insurance to value is critical for both the insurer, which needs to collect adequate premiums, and for the customer, who needs building limits that will cover reconstruction costs and avoid possible problems with falling short of coinsurance requirements.

Geocoding. This provides the latitude-longitude coordinate of a risk's address. The underwriter can then determine how far the risk is from potential hazards such as brush fires, terrorist targets, bodies of water, wind-exposure areas and earthquake fault lines. This is important for underwriting individual risks and crucial for controlling spread of risk. (For instance, after a hurricane, insurers often find out too late that their insured properties were too concentrated in a vulnerable area.) Geocoding allows insurers to determine and control their spread of risk. If the geocoding information is stored in a data warehouse or data mart, managers can readily find out the geographic concentration of business and determine the exposure of their book of business to major hazards. Geocoding also helps insurers make more informed decisions about catastrophe reinsurance.

Risk-specific data. This includes data on specific companies and business owners, including payroll receipts, financial data, bankruptcy records, liens, judgments, creditworthiness and whether the applicant is on the terrorism watch list. Much of this is of obvious value: payroll has to be correct for calculating workers' compensation premiums, and most insurers and agents are wary about dealing with companies that are so financially shaky they may not be able to pay premiums on time or even finance their premiums.

But some of the value is a bit more subtle. In a small business, the owner is the business to a large extent, and the underwriter would want as much relevant personal data on the owner as possible. For example, when insuring a serf-employed trucker or small delivery business, the underwriter would want to know about the owner's personal driving record, readily available through a motor-vehicle report. How the applicant drives and maintains a vehicle will indicate much about how he or she operates the business. Personal credit ratings and bankruptcy records are vital for the same reason.

Integrating and Using the Data

System integration is the key to using third-party data efficiently. If the policy management system is a contemporary one, with a service-oriented architecture built to accept outside data and work with external systems, integration is simple. These systems use extensible markup language (XML) and are designed for messaging--the ability to take data in and transmit outputs. With older systems, however, integrating external data sources presents greater challenges.

Easy access to data is also key. For instance, an integrated desktop makes the user's life easier. It would be unnecessary for the underwriter to switch to a different desktop to look up external data; access is through the underwriting system's desktop. Alternatively, data services can run in the background automatically.

As an example, assume the underwriter is rating a vehicle. The underwriter enters the vehicle's year, make, model, cost and vehicle identification number. When the VIN is entered, it immediately triggers an inquiry to a third-party database, providing real-time feedback to the user. If the gross weight or cost new is off, the underwriter can correct it immediately. This is much more efficient and less costly than accepting the risk, doing a physical inspection and finding out that vehicle weight (or payroll figures or building valuations) are wrong and changing the records weeks later.

The outside data systems can run in the technology background, flagging the underwriter only if there's a specific concern that must be looked at. Under the USA Patriot Act, insurers and other financial firms must report suspicious transactions to the federal government. This requires clearing applicants against the terrorist watch list. But it's not quite as clear-cut as it seems; multiple watch lists exist and a completely innocent person could have the same name as a notorious terrorist. The outside database can flag the underwriter that the name is on a watch list. The insurer can conduct more investigation to determine if the applicant is truly the same person who is on the watch list, and if so, follow federal reporting regulations.

Similarly, if geocoding flags a major-hazard risk, the underwriter could reject it or raise the rate appropriately.

An intelligent underwriting system can do more with more data. This requires adding analytics, sophisticated mathematical models that help predict the losses likely to be incurred by a risk. Analytics can transform data into both a descriptive instrument that shows where a company has been, and, more important, into a predictive instrument that tells it where it should go. Integrating the model into an underwriting system to combine data and analytics will boost underwriting efficiency and profitability throughout the organization.

Key Benefits

The cost of third-party data has fallen substantially over the years. The improved underwriting it promotes easily pays for the modest costs of tapping into the databases.

Automating much of the underwriting process helps boost productivity. Rather than manually reviewing each routine decision--an expensive proposition in small and midsize commercial lines where premiums are usually modest--underwriters can concentrate on dealing with the exceptions.

With greater intelligence and more data, underwriting could be pushed further out to the user community. When underwriting rules, analytics and outside data sources are incorporated in the policy management system, agents and brokers could quote and underwrite the majority of smaller commercial risks, while company underwriters handle only cases identified as exceptions.

Today, insurers of all sizes and types, from small regional carriers selling through agents to national direct writers, can take advantage of third-party data and the latest underwriting technology, and underwrite small and midsize commercial risks as diligently as they write individuals and large businesses.

Key Points

* Data needed for underwriting fall into three categories: validation, geocoding and risk-specific.

* Outside data systems can run in the technology background, flagging the underwriter only if there's a specific concern.

* Underwriting systems can do more if they have sophisticated mathematical models that help predict the losses likely to be incurred by a risk.

Contributor Dorrie Pighetti is vice president and chief insurance officer with Hartford-Conn.-based Insurity, a provider of property/casualty polity administration software and outsourced services.