When asked for the biggest challenge facing the Czech automobile insurance industry, Peter Jedlicˇka, PhD, doesn’t hesitate. “Bodily injury claims are growing disproportionately compared with vehicle damage claims,” says Jedlicˇka, team leader of actuarial services for the Czech Insurers’ Bureau (CIB). CIB is a professional organization of insurance companies in the Czech Republic that handles uninsured, international, and untraced claims for what’s known as motor third-party liability. “Bodily injury damages now represent about 45% of the claims made against our members, and that proportion will continue to increase because of recent legislative changes.” One of the difficulties that bodily injury claims pose for insurers is that the extent of an injury is not always predictable in the immediate aftermath of a vehicle accident. Injuries that were not at first obvious may become acute later, and apparently minor injuries can turn into chronic conditions. The earlier that insurance companies can accurately estimate their liability for medical damages, the more precisely they can manage their risk and consolidate their resources. However, because the needed information is contained in unstructured documents such as accident reports and witness statements, it is extremely time consuming for individual employees to perform the needed analysis. To expand and automate the analysis of unstructured accident reports, witness statements, and claim narratives, CIB deployed a data analysis solution based on Dell Statistica Data Miner and the Statistica Text Miner extension. Statistica Data Miner offers a set of intuitive, user-friendly tools that are accessible even to nonanalysts. Application Case 5.1 Insurance Group Strengthens Risk Management with Text Mining Solution The solution reads and writes data from virtually all standard file formats and offers strong, sophisticated data cleaning tools. It also supports even novice users with query wizards, called Data Mining Recipes, that help them arrive at the answers they need more quickly. With the Statistica Text Miner extension, users have access to extraction and selection tools that can be used to index, classify, and cluster information from large collections of unstructured text data, such as the narratives of insurance claims. In addition to using the Statistica solution to make predictions about future medical damage claims, CIB can also use it to find patterns that indicate attempted fraud or to identify needed road safety improvements. Improves Accuracy of Liability Estimates Jedlicˇka expects the Statistica solution to greatly improve the ability of CIB to predict the total medical claims that might arise from a given accident. “The Statistica solution’s data mining and text mining capabilities are already helping us expose additional risk characteristics, thus making it possible to predict serious medical claims in earlier stages of the investigation,” he says. “With the Statistica solution, we can make much more accurate estimates of total damages and plan accordingly.” Expands Service Offerings to Members Jedlicˇka is also pleased that the Statistica solution helps CIB offer additional services to its member companies. “We are in a data-driven business,” he says. “With Statistica, we can provide our members with detailed analyses of claims and market trends. Statistica also helps us provide even stronger recommendations concerning claims reserves.” Intuitive for Business Users The intuitive Statistica tools are accessible by even nontechnical users. “The outputs of our Statistica analyses are easy to understand for business users,” says Jedlicˇka. “Our business users also find that the analysis results are in line with their own experience and recommendations, so they readily see the value in the Statistica solution.”
Questions for Discussion
1. How can text analytics and mining be used to keep up with changing business needs of insurance companies?
2. What were the challenges, the proposed solution, and the obtained results?
3. Can you think of other uses of text analytics and text mining for insurance companies?