Travel insurance fraud is a soft target for fraudsters – claim processes do not match the level of verification/investigation found in other insurance product lines. It is a common occurrence, and incidence is rising each year.
More travelers world over have submitted fraudulent claims to their travel insurers as they look to recoup the cost of their holiday. Research by the UK’s Direct Line Travel Insurance has estimated that nearly eight million people in the UK alone have made a claim on their insurance that was fraudulent in some way, with five per cent of those people fabricating their claim completely, and 15 per cent inflating the value of their claim. And that number is rising all the time. We are looking at a scam that’s costing the industry millions.
There are several reasons why travel insurance fraud is a soft target, and a unique problem compared to other types of insurance fraud:
- The incentive for travel insurance fraud is not an external issue.Fraudsters are usually trying to recoup/offset the cost of their trip.
- Potential fraudsters are under the impression that travel insurance fraud is not a serious offence like other types of insurance fraud. Cheating a travel insurer is seen as ‘stretching the truth’ more than a criminal offence.
- Travel insurers have tighter margins and lower premiums – harder to invest in an elaborate, lengthy fraud detection process.
- Due to the competitive nature of this business, less information is obtained from customers It is also much easier to obtain a travel insurance policy.
- Incidents (upon which claims are made) usually occur in uncontrolled/variable and undocumented environments (for example, overseas). The variables are also much harder to validate(for example, the contents of a lost bag).
- The insurance industry’s anti-fraud focus has so far mostly been in auto or personal injury. Fraud detection in travel insurance is not so well formulated.
Travel insurance fraud usually involves investigation by cognitive interviewing, front-end claims, assistance and medical teams, back-office specialist investigation staff, overseas agents, etc.
As yet, data analytics did not seem to be used much for fraud detection in travel insurance.
Until now. That is.
Starting in mid-April this year, a specialist team from Saama flew to the client site to tackle an interesting but uniquely challenging problem: catching fraudsters in the worldwide travel insurance market.
Saama built a business flag model. With statistical methods, and borrowing from their significant experience with recent fraud detection projects in the insurance domain, they were able to identify five crucial flags. Then, an ensemble of predictive models were designed using machine language methodologies. This included penalized logistic regression, gradient boosting machines with under sampling, and a deep learning auto-encoder for anomaly detection. All approaches used stratified cross-validation to optimize internal model parameters and customized business metrics.
A beautiful and functional visualization and user interface was designed. In addition, the potential for incorporating government travel alerts, travel warnings, individual twitter handles into predictive models, or finding newsworthy events using anomaly detection on Google search trends data.
Truth be told, It was mind blowing to see data science be used in this relatively unknown domain.
Learn more about Saama’s Fluid Analytics for Insurance and what we can do for you.