Last week, I introduced the idea of an anti-fraud program for insurers. Today, I’ll talk about some of the analytics that can be implemented as part of these initiatives.
Business rules and anomaly detection
Typically, the first line of defense in fraud screening is the use of business rules and anomaly detection. Algorithms of known types of fraud are applied against claims, with the aim of identifying specific activities. However, these techniques are limited in dealing with a large number of variables. Further, sophisticated fraud rings are quick to discover a carrier’s business rules and will craft fraudulent claims that slip under the radar.
The next line of defense is analysis of unstructured data. Text mining can be a very effective tool, especially given the large quantities of text-based information that is generated during the claim process. These include resources like e-mails, customer service, calls, interviews, adjuster notes and other sources.
Social networking and link analysis
Social networking and link analysis can also be used to detect fraudulent activity. These analyses can identify relationships between and among claims, identify patterns and highlight possible links to suspicious activity across large numbers of claims. Tools are available to map activity across a selection of customer claims, providing a graphical representation of how activities are connected. These tools are valuable in identifying fraudulent claims from organized crime.
These are only some of the tools and analytics that compose an insurer’s anti-fraud tool box. It is important to be cognizant of any and all information sources, and to apply analytics to that information to look for patterns. These patterns can be applied to the present, to look for links between claims and highlight suspect activities. They can also be synthesized into predictive modeling, to try to predict fraudulent claims in the future.
Next week, I’ll talk about how to measure the success of an anti-fraud program.