Insurance companies have long been seen as easy and, in some cases, acceptable target for committing fraud. Globally, insurance fraud is the second largest and most financially lucrative criminal enterprise after illegal narcotics.
For that reason, whenever insurers put counter-fraud defences in place to stop the progression of organised crime, fraudsters will find a way around them. Criminals are continually evolving and the speed at which they develop new behaviour is impressive. In order to keep up, insurers need to maintain technical agility.
Technological advancements in fraud detection, if deployed correctly, enable insurers to tackle the continual threat they are facing, and help prevent fraud losses from affecting their combined operating ratio.
Zero-tolerance vs. customer satisfaction
Just like in any other sector, new customers are costly for insurers to come by, yet it’s far easier than it ever has been for consumers to switch suppliers. Customer retention is of utmost importance.
Many executive teams are taking a zero-tolerance approach to how fraud impacts genuine customers. Claims have to be resolved quicker, often driven by regulation, and customers are increasingly expecting faster claim resolution and payment.
Currently a percentage of claims normally have to be resolved within 24 hours, or five working days. The challenge insurers face is they know that somewhere between 10 and 20 percent of claims contain fraud – either opportunistic or criminally organised. But how do you find that one-in-five, and at the same time not slow down payment to your genuine customers?
In select cases the market is moving toward low-touch or touchless claims and straight-through processing, reducing the overall cost of claims management. As the market moves so will fraud and it’s important that companies keep in advance of the fraudsters.
The challenge of both validating genuine customers and spotting fraudsters requires insurers to move faster than ever before.
Fraud and high risk detection at underwriting
While claims fraud detection remains a vital capability for insurers and its importance cannot be understated, recent emphasis has been placed on fraud detection at point of application – screening people as they come through the door.
The boom in online channels – offering the ability to communicate with insurers in ways consumers couldn’t before – has had a significant impact on the need for policy screening. The control is now very much in the hands of the consumer, rather than via a broker.
Before, individuals would go to see a broker and buy the policy they recommended. Now, there are multiple online channels, aggregators and price comparison websites. Consumers can shop around, change their details slightly here and there – there’s little personal contact.
That loss of personal contact over the last five to 10 years means that insurers have had to increase the sophistication of fraud detection techniques.
Traditionally, fraud checks at policy inception have taken the form of “data washing” i.e. has the carrier seen this applicant before and are they suspicious? This can generate a lot of false positives.
However, sophisticated organised fraud rings and emerging modus operandi mean network analysis and machine learning need to be employed to detect suspicious activity at the point of claim and policy inception.
High-volume, low-latency detection is truly important here. Dealing with high volumes of data quickly is vital.
The scale of the data processing that insurers need to be capable of is such that they can’t do it with existing technology. In the past, insurers would process a lot of this data on traditional relational databases, scripting languages and analytics tools, but they need to be able to scale the amount of processing they’re doing on much cheaper hardware.
Hosted solutions and clusters of commodity-based hardware can be used to provide greater processing power and the flexibility to dial capacity up or down just when it’s needed.
Insurers hold a large amount of data that can be harnessed for effective fraud detection. The growth of artificial intelligence (AI), the Internet of Things (IoT) and smart devices has increased the amount of data insurers have access to. The challenge is how to apply that wealth of data to streamline claim resolution and fraud checks, thus enhancing efficiency and customer satisfaction as a result.
Over and above the information an individual insurer has access to, there is general recognition of the value that third party data can bring to the fight against insurance fraud.
There’s intel from the Insurance Fraud Bureau, the Insurance Fraud Register, information from credit bureaus, publicly available information such as the electoral roll, unsatisfied County Court judgments and bailiff searches.
All of this data, if used effectively, can be used to prove the identity of a customer, and aid in indicating risk.
When it comes to data use, insurers have historically been ahead of other industries, and have used data from outside their organisation. The challenge is the maturity of their approach to then using that data, and the analytics they’ve performed on it.
Collecting data for the sake of it is costly. For fraud detection purposes, insurers need to be performing appropriate analytics and using the right tools to find that proverbial needle in a haystack at the lowest possible cost.
To some, the insurance sector may traditionally have been seen as behind others technologically. However, in the area of data sharing, particularly in the UK, the industry is ahead of the curve. The IFB, part of the Association of British Insurers, is a fantastic example of this.
Challenges still exist internally with sharing intelligence such as tip-offs, known fraud lists or internal black or grey lists across different parts of the business. In the fight against fraud it is paramount that intelligence is managed carefully and appropriately, but also disseminated in a secure and timely manner to aid investigations.
The problem is less about data siloes in this case – it’s about intelligence siloes. The fraud team knows a lot about what’s going on in the claims world, but they often have no process or way to effectively share that intelligence with the systems that are in place to detect fraudsters earlier in the lifecycle, including at policy inception.
Agile deployment of technology
Insurers need to deploy their fraud detection technology with agility in mind.
The conversation they need to have can be summarised as: “I don’t know what problem I’m going to need to solve yet, so let’s architect our solution in a way that means I can solve problems I’ve not yet anticipated, with volumes and types of data that I’ve not yet seen.”
Applying advanced analytics and machine learning to the vastly increasing data volumes and appropriately managing data and intelligence within and across organisations is absolutely fundamental to detecting fraud quickly and efficiently.
Insurance companies need to build their processes, train their staff and have the right connections between departments to ensure that they’re mitigating every threat.