With Aviva reporting a 13 per cent rise in fraud claims last year, Colin Bristow, fraud & anti-money laundering specialist at SAS UK & Ireland, looks at the techniques insurers can use to stem the rise.
It’s sometimes difficult to know if fraudulent insurance claims are really rising, or whether companies are just getting better at detecting them. Most likely, it’s a combination of the two.
Certainly, the techniques fraudsters deploy, including their use of technology, are becoming more sophisticated all the time. Research from Aviva reveals high numbers of ‘ghost brokering’ cases last year – where policies are bought using details customers have been misled into handing over, often on social media, or where fake documents that look like they’re from legitimate insurers are sold on to customers. Indeed, of the 20,000 motoring fraud application cases Aviva detected, ghost broking accounted for 15 per cent of them.
Fraudsters are also quick to exploit events like the pandemic, the whiplash reforms and the current cost of living crisis for their own financial gain.
For instance, there was a 10.7 per cent spike in motor injury claims ahead of the new whiplash rules coming into force in May last year, according to Aviva. Since then, the focus seems to have shifted to fraudulent vehicle repair and credit hire, which jumped by 13 per cent in 2021.
Trying to outwit the fraudsters is an ongoing challenge, especially as the volume and complexity of the investigation increases. Anti-fraud teams face mounting pressure to quickly root out false claims, so that legitimate ones can be processed quickly and painlessly and customers aren’t unfairly penalised.
Companies’ bottom line and reputation depend on whether or not they achieve this – yet it remains a resource-heavy task, requiring more staff. This isn’t cost-effective, nor is it scalable. They need tools that streamline fraud detection processes and allow them to proactively monitor trends and changes in fraud activity, not just respond to existing risks.
Data analytics have already transformed fraud detection methods. According to our research, conducted in partnership with the Association of Certified Fraud Examiners (ACFE), 64 per cent of organisations say data analytics enables them to review higher volumes of transactions. Other benefits include improved accuracy (i.e. reduced false positives); automating tasks to increase efficiency; and identifying anomalies in a timely way.
It’s highly unlikely that manual methods alone would have enabled Aviva to uncover 11,000 instances of claims fraud in 2021, worth more than £122m, and investigate more than 16,700 claims under investigation for suspected fraud.
Our research shows that around two-thirds of insurers currently still use more traditional analytics methods such as exception reporting or anomaly detection, and over half have automated monitoring of red flags or violations of business rules in place. These methods will remain the most widely-used, at least over the next couple of years, but as digital solutions advance, they’ll be complemented by newer anti-fraud techniques.
The proportion of organisations using artificial intelligence (AI) and machine learning (ML) in fraud detection is expected to almost double over the coming years, from 13 per cent currently to 25 per cent, and predictive analytics and modelling is set to rise by 22 per cent.
Applying AI and ML techniques, in particular, can add significant value to fraud investigations. For a start, it would allow insurers to draw on more data sources to develop a clearer picture of what’s happening, without needing more people to process it.
AI and ML have been shown to reduce false positives so customers with genuine claims aren’t unfairly penalised. You can also analyse data from across the business – motor, homes, travel and so on – to identify new and emerging trends more easily.
We found that three-quarters currently use internal structured data in their anti-fraud tests (for example, from their claims management software or CRM) – but developments in digital solutions mean you can layer it with other data sources and spot vital clues.
Around 43 per cent already analyse public records, while 32 per cent use government or law enforcement watchlists. Another 30 per cent look at their own unstructured data, like phone calls, emails and instant messages to check for potential fraud.
Real-time network analysis allows insurers to accurately map the connections between drivers, passengers and body repair shops in the case of automotive fraud – adding clarity and context to investigations. Data analytics is also helping firms to proactively monitor more risk areas, including fraud by customers, third parties, staff and contractors, and money laundering.
As powerful as these data analysis tools are, it’s worth remembering that none are a silver bullet by themselves. Insurers need to be able to quickly apply a range of data analysis methods, and automate aspects of risk ranking. Being able to detect and rank potential fraud cases in order of business risk means you can focus your limited resources on the most damaging ones.
Clamping down on fraud requires insurers to be more vigilant than ever in today’s climate.
As the cost of living crisis begins to bite, customers may well be pushed towards ghost brokers offering ‘too-good-to-be-true’ deals online, or be tempted to falsify/inflate claims.
We should remember too that insurance fraud has far-reaching implications, beyond hitting companies’ bottom line and pushing up premiums for customers. It is not a victimless crime. Scams are often linked to serious organised crime, including ‘cash for crash’ which puts people at risk, and will prevent insurers delivering on their ESG strategy.
Finally, the world of insurance is more competitive than ever – so the ability to detect fraud, while ensuring legitimate claims are paid promptly, puts insurers at a commercial advantage. Not only can they control the costs of claims processing, they could reduce sales and marketing costs because customers are more likely to stay and recommend the company.