This article was re-printed from National Underwriter, June 1996.

It has been provided courtesy of Business Guidance Systems Inc.

Detecting Fraud With Insurance Data Warehouse

By Harold Snider And Tam Seto

We all pay for the cost of crime, and preventing it is much more appealing for insurers than accounting for it after the fact. Insurance losses related to crime and abuse are factored into companies' rates as a cost of doing business.

Relatively few instances of fraud affect the balance of the companies' customers. Insurers have implemented sophisticated and powerful computer systems to try to accurately identify the losses as soon as possible after they happen, and factor them into their rates through timely accounting of the losses of those few.

Recovery of the losses after the fact does not happen as quickly as the loss itself, since new rates must cycle through the natural course of business as new policies occur and old ones are renewed; and you don't just add in the real costs, they have to be factored into the rates carefully, after considering competitive and shareholder concerns as well. Powerful actuarial resources are in place to forecast and predict the necessary reserves to protect the insurer against these few potential losses from crime and abuse. This is really a form of accounting for crime that is expected to happen. What about preventing it before it happens?. A recent study revealed that 10-15 percent of insurance premiums fund the North American $40 billion insurance fraud tab, not including the accompanying investigation expenses and legal fees.

Fraudulent claims are not only very expensive, they are also one of the most frustrating and aggravating elements of the insurance industry. Conventional wisdom in the industry states that 10 to 20 percent of all indemnity is fraudulent. The percentage of claims which are detected or denied ranges from 1 to 5 percent, suggesting improvements are to be gained. This gain heads straight for profit line of the insurer's balance sheet. For many of the major insurers this gain means millions to hundreds of millions of profit not being pursued. Even a small improvement of a few percentage points is significant, and the potential for improvement is much greater.

Reducing or controlling this significant amount is worthy of proactive investment by insurers. Just as public law enforcement agencies continually search for tools and techniques to help them prevent crime, Special Investigative or Security Units of insurers can also use a helping hand. Tools are important, and the raw material is under their noses. In addition to currently available external industry claims data bases, insurers have a very powerful resource within their own computer data centers, their own operating data.

For many insurers, fraud investigation is handled in a responsive and reactive manner, When claims administrators are suspicious of a claim, they inform their supervisor. The supervisor reviews the information, and if the supervisor feels the claim warrants investigation, it is forwarded to the SIU or Security Unit, The investigative/security unit reviews the information to determine whether to accept the claim for investigation, The investigators have at their disposal specialized investigative techniques and artificial intelligence software products to perform their research.

However, this inspection is focused on specific individual cases or at a small number of claims. Any proactive initiative is completely dependent on the experience or inquisitiveness of the person initially processing the claim(s). Fraudulent intentions are difficult to detect on individual basis, unless blatant. Trending insights are not available. Dormant exposures are not visible, Abusive patterns are not obvious.

On the other hand, timely and easy access to enterprise operational data which is integrated and complete will enable the fight against fraud to be powerful and more effective. An insurance data warehouse which contains organized detailed historical data will provide fraud fighters with a very powerful weapon.

Uncovering fraudulent claims requires extensive data gathering and analysis. Often the information is difficult and time consuming to obtain. It has to be manipulated into an environment that accommodates an analytical process - a data warehouse. Further, the information is time sensitive in many cases, and there is the need to acquire it in an efficient manner. There needs to be sensitivity to the confidentiality of the data collection process, and this is difficult in the environment currently in place in most organizations.

Finally, the investigation & security staff has a requirement to ask their questions and receive answers in a timely manner, while they are in the midst of a certain thought process. It is a heuristic process. Answers lead to more questions, and so on. How can a data warehouse help fight fraudulent claims or exposures? The ability to identify or detect an investigative path and to follow this path is a primary benefit. Let's explore some examples.


Automobile Insurance:

As automobile insurance premiums represent over $9OB in the United States, insurance fraud is a particular problem for auto insurers. Fraudulent claims present themselves in various forms like:

The current reaction capability of investigators limits them to pursuing after the fact or evaluating the situation in isolation. The ability to apply the lessons and profiles gained from other policies is very difficult or nonexistent. However, with an integrated data warehouse which is easily accessible, easy to use, and flexible enough for ad-hoc querying, fraud investigators can activate a proactive approach to combating these costly claims through earlier identification and prevention. Such a data warehouse would typically contain:

In a recent exercise, a major P&C insurer gained an ability to explore its own data integrated in such a data warehouse. In the first step, the insurance professionals were asked to produce a list of policyholders who made claims over a particular amount within the last two years in a particular state. Then the investigators asked for the results to be grouped by Vehicle Identification Number (VIN). What they observed was alarming to them; there were multiple situations, several thousand vehicles, where multiple active policies were covering a common VIN. Armed with such information, the insurers' SIU went into action and ultimately saved the insurer millions of dollars in repossessed claims or unnecessary exposures.

Then the investigators looked for claims for recently insured high priced vehicles involved in rear-end collisions with older vehicles, a favorite scenario for staged accidents. As well, they uncovered several cases where they had paid a loss for stolen, unrecovered cars which were now insured by another of their companies, in some cases by the original owner.

With such an integrated data warehouse, this exercise was completed in half a day where previously it was impossible or required weeks of efforts involving SIU and information technology resources. In this way the company took a major leap forward, from rationalizing buried rate increases, to cutting their losses and increasing their competitive edge. This project, which cost almost nothing, was completed in months and returned millions within a matter of one month of utilization.

Insurers have experts on claims processing and experts on the investigation of fraud, but no experts on the detection of fraud in this sense. Investigative expertise and experience combined with flexible and timely access to enterprise wide operating detail arm insurers with proactive fraud fighting capability. A rapidly constructed enterprise insurance data warehouse will provide insurers with a proactive ability to keep claims costs down for a strategic ability to price their products more profitably and competitively. This same data warehouse can also service other needs of insurers, like product development and target marketing for even stronger market position.

Mr. Seto is Initiative Manager, Insurance Industry Decision Support and Data Warehousing Solutions for Tandem Computers Inc. in Cupertino California.

Mr. Snider is President of Business Guidance Systems Inc. and a Business Consultant for Decision Support and Data Warehousing Solutions for Tandem.


Copyright Notice