The first challenge comes from companies' endeavours to expedite their claims processing procedures and - at least in the case of frequent claims - convert them to a low-file or non-file basis. The logical outcome of this development is that claims are being recorded and processed at call centres, which are supposed to decide upon the further course of action over the telephone.

At the same time, there is the need for insurance companies to master the problem of fraudulent claims. Estimates, which should obviously be interpreted with care, assume that in Germany approximately 10% of all traffic accidents in the motor third party liability (MTPL) sector have a fraudulent background. This figure relates to the area of professional crime, in other words, to arranged and provoked accidents. This increases insurance companies' loss burden by approximately DM2 billion per year. However, fraud can only be effectively combated if fraudulent claims are identified early enough to prevent the payment of compensation in the first place.

ICE approach

It is practically impossible to tackle the task outlined here by traditional means. However, innovative software enables the empirical knowledge of fraud experts concerning typical fraud patterns and corporate policy procedures for handling such cases to be stored in insurance companies' IT systems. In cooperation with five significant German primary insurers and the software provider Inform GmbH, GeneralCologne Re's claims department developed an IT system for the automated identification of fraudulent claims in MTPL insurance called ICE (Intelligent Claims Examination) /MTPL.

When claims reports are entered. they are automatically scrutinised in terms of the likelihood of fraud on the basis of information on the policyholder and the insured vehicle, accident details and the other party involved in the accident. The primary objective is to reliably identify unsuspected “black sheep” and, even in cases where the circumstances are not clear, to subject as many cases as possible to rapid processing procedures by considering incriminating and exonerating factors. Only suspicious cases (ie. where the system calculates a high probability of fraud) are then submitted to the fraud experts for further assessment.

Such an automated fraud identification system offers major advantages.

  • The prompt identification of fraudulent circumstances results in a direct reduction in the loss burden. According to initial estimates, the savings potential is a substantial percentage of the MTPL premium.

  • The rapid processing of as many claims as possible optimises client satisfaction and fully exploits the savings potential inherent in rapid claims processing.

  • The system supports members of staff who are responsible for recording and initially processing claims so they can do the work of qualified and experienced claims assistants.

  • Tailored towards the capacity of the fraud experts, such a system enables the appropriate number of claim notifications, which involve the greatest probability of fraud, to be filtered out. This means that the fraud experts are spared the time-consuming task of pre-sorting suspicious cases and can focus their expertise on the most likely cases.


    The first module for MTPL business is currently being brought into productive use by the first clients. A second module for motor own damage insurance (ICE/MOD) is in the last stage of development. Development work on two further modules for general liability and private property is planned to start this autumn and should be completed in the first quarter of 2001.

    Overall, the project has been very successful in Germany. It is assumed that, by the end of this year, insurance companies representing about 50% of the market premium will have bought one or all of the ICE modules. Furthermore, GeneralCologne Re has started to transfer and adapt the system to other European markets.

    Eberhard Fähnrick is a member of GeneralCologne Re's global claims department and the creator of its electronic fraud detection software.

  • Topics