(Image source: Valen homepage.)
Valen Analytics, a provider of proprietary data, analytics and predictive modeling for property/casualty insurers, has announced the latest in its suite of predictive models with the launch of the Unavailable Loss History Model. Describing the new offering as “first-to-market,” Valens says the new model enables accurate scoring of insurance policies that have no information on loss history.
Though prior loss experience has traditionally been one of the key parameters in evaluating the quality of the risk at quoting, insurers can score up to 70 percent of small premium policies without any loss information using models that aren’t built for these types of policies, a Valen statement notes. Valen’s model, which the vendor says has gone through generations prior to release, is designed to delivers\ highly accurate predictions for insurance policies without any loss history information by employing a combination of third-party and synthetic variables derived from the Valen Data Consortium. The model is built and tested against a large and granular data set of approximately 650,000 policies presenting $7.6B in premium, the vendor reports. By leveraging that data set with 10 years of loss experience, Valen says, the new model is finely tuned to answer the question “How do similar policies perform?” as an indicator of risk quality.
“With this latest solution, we are entering the next stage of industry innovation in predictive analytics,” comments Kirstin Marr, President of Valen Analytics. “Being able to deliver valuable insights despite the absence of a key piece of information like loss history is a true testament to the power of partnering around data. As insurers pursue the small commercial market, the ability to accurately assess risk, while collecting less information on the application is crucial to gaining market share. We are confident that we will continue to develop such impactful solutions as our Consortium scales.”
Growing Data Set
Valen reports that it recently conducted a study that demonstrated that synthetic variables developed with the vendor’s Consortium data demonstrated up to 13 times more predictive power than variables built with policy-only data. Synthetic variables are built from computations of more than one variable and leverage the breadth and depth of experience contained in the Consortium. The vendor reports that its Valen Data Consortium—which is composed of data from dozens of third-party data sources and more than 60 insurers—has significantly grown in the past year, with 57 percent growth in Workers’ Compensation and 266 percent in Commercial Auto. “This momentum allows Valen to better serve insurers with more accurate predictive insights needed for real-time decision making,” the Valen statement says.