For insurers that commit to analytics, Valen says it is more than just adding tools, but also a change in organizational philosophy. “Implementing sophisticated analytical tools within underwriting requires a commitment to becoming a more analytically driven organization,” explains the report.

C-level commitment represents “the most important first step,” Valen says. CEOs, the report adds, typically want to know:

• That the predictive model works.
• That predictive analytics will drive meaningful and demonstrable benefits—for example profits, loss-ratio improvement and pricing accuracy.
• The implications to the organization—IT impact, changes to the underwriter workflow, how underwriters will be trained and implications to key stakeholders such as agency relationships.

Once the organization buys in, Valen says subsequent considerations are what data assets to use to build a robust predictive model, how information will be consumed within the organization and incorporated into the underwriting workflow, and how predictive scores will work “in synergy with the expertise your underwriters bring to the table.”

Regarding the data to be used, Valen notes that selection bias can occur when predictive models are built solely on a carrier's own data. “Your company has a risk appetite along with risk selection and underwriting guidelines that help define your profile of business,” says Valen. “Therefore, if you build a model on your data alone, the dataset will only include those policies that your underwriting practices selected and, by definition, does not represent the entire market.”

Valen notes that carriers may use third-party data or leverage data consortiums to supplement their own policy data.