Tackling Insurance Fraud

Insurance fraud is a significant and costly problem for insurance companies worldwide. According to the Association of British Insurers (ABI) £1.1 billion in insurance fraud was detected in 2022 in the UK and it’s estimated only a portion of fraud is detected. Undetected fraud may double or triple the actual losses.

Insurance fraud detection can be undertaken in many ways.

Detection on Inception

Detection on Inception is great for stopping known risks or opportunistic fraud. Rules can flag known risks while external sources can validate a person’s identity, address or even credit score.

“For every piece of fraud detected it’s estimated that double or even triple that amount goes undetected. This is where the dangerous fraud and organised crime lies.”

Post-Inception and Continuous Monitoring

For every piece of fraud detected, it’s estimated that double or even triple that amount goes undetected. This is where the dangerous fraud and organised crime lies.

Post-Inception Fraud Detection focusses on emerging fraud signals, policyholder behaviour over time, or suspicious claims patterns.

Continuous Monitoring focusses on model tuning, fraud leakage quantifications and identifying missed frauds.

This is where GRAPHT comes in.

Dashboards can identify trends but lack the detail to identify actual fraud. While Network visualisation has been proven to detect by visualising behaviour and patterns as networks.

Legacy network visualisation products have always been limited in scale. This means that intelligence/investigation teams must concentrate on small time series or known fraudulent patterns, by writing queries to narrow down to a manageable size. But what if the fraudulent pattern is not known?

Take a large insurance company who might process over 350,000 claims annually. They could process in excess of 7,000 claims a week.¹

Now let’s assume that each claim has data points linked to it, the claimant, an address, email, phone number, maybe a vehicle. Each claim + its linked data points could generate 70,000 items to visualise, considerably more than what the current legacy products can visualise.

What if you want to analyse more than a weeks worth of data?

Introducing GRAPHT

GRAPHT allows intelligence teams to view more of these networks (100 times more than legacy network visualisation products²), allowing these fraudulent networks to come to the forefront in 3D visuals with analytical features that simply aren’t available anywhere else.

It’s estimated that undetected fraud costs a large UK insurer between 100 - 120 million a year.³

Can your organisation afford to not use GRAPHT?

ABI reported that UK motor insurers handled 2.4 million claims in 2024. Assuming motor claims account for 65% of total general-insurance claims, the total UK volume is 3.7 million claims per year. Large insurers with 10% market share would therefore process 370,000 claims annually. Each claim may link to multiple data points (claimant, address, vehicle, email, etc.), generating tens of thousands of items to visualise across the network.

Estimated from 10,000 items being usably visualised on a modern PC using legacy network visualisation software. With GRAPHT usably visualising 1,000,000 items on the same machine through the browser.

Detected insurance fraud in the UK was estimated at £1.1 billion in 2022 (ABI). Studies suggest a similar amount of fraud goes undetected each year, implying total UK fraud of roughly £2.2 billion. Large insurers with 10% of the UK market would therefore face a total fraud exposure of £220 million annually, of which approximately half (£100–120 million) is detected, with a similar amount (£100–120 million) estimated to be undetected.