Fraud - voice stress analysis: Liar, liar


With fraud costing insurers £1.8bn in 2009, Daniel Dunkley explores the validity and effectiveness of voice stress analysis technology, once hailed as a 'silver bullet' for the industry.

Liars are rarely adored in any walk of life, but one of the most financially susceptible industries to people's little white lies — and the far more serious ones — is the insurance industry. According to the Association of British Insurers, fraud cost the industry £1.8bn last year, and anecdotal evidence suggests the recessionary environment will only see this figure rise further in 2010.

Luckily, with the advent of the technological revolution, insurance companies are able to call on more sophisticated methods in their attempts to thwart those attempting an opportunistic and fraudulent claim. Among these methods is 'voice stress analysis' — or lie detector technology as it has often been referred to.

VSA relies on technology that is said to identify psychophysiological stress responses present in the human voice when a person suffers 'stress', and other suspicious emotions in response to a question. The fear caused by a liar's exposure can cause a 'high stress' voice signature, which can then be traced by VSA.

The subject of much fanfare in the insurance market back in 2004, VSA was quickly heralded as the 'silver bullet' by some companies keen on rooting out opportunist fraudsters for good. Digilog is the most prominent VSA software provider operating in the UK insurance industry, and works with a number of providers in rolling out this software for claims handlers. The firm is even working with the Department for Work and Pensions to scan benefit claims for possible discrepancies in 20 local authorities.

As with any new technology, questions are posed over the actual contribution this analysis software makes to fraud detection, and opinion — as so often in the insurance industry — is divided.

Only last month, Post revealed the deal that saw Mike India 5 take over the motor investigation arm of Teceris, including a 10-year deal for the exclusive outsourcing of Digilog's software in the insurance space ( As far back as 2002, major brokers and insurers, including RSA, were understood to be using Digilog. But in March 2005, RSA withdrew its support for the technology and other insurers, in particular Aviva, were sceptical from the start.

A total of 15 insurance companies, including LV's Highway business, Esure, Provident and Chaucer, still utilise the technology as part of their counter-fraud strategies. There is great interest in how useful, reliable, and cost efficient the technology is. So, will VSA be insurers' best friend for years to come? Or has the original fanfare caused by its arrival proven to be a false dawn in the battle against fraud?

A new era

Motor insurer Provident doesn't seem to think so. The insurer seems mightily impressed by its work with Digilog, and insists the technology — despite its imperfections — is a hugely useful tool in the fight against opportunistic fraud. The motor insurer, which has been using the technology for five years, operates VSA alongside a list of cognitive interviewing questions posed by its staff, and cites Digilog as heralding a new era for claims efficiency. Provident currently employs VSA in theft and fire claims, as well as to uncover "exaggerated personal effect claims". Andy Haslam, claims supervisor at the insurer, says 'pre-Digilog' it was only refuting 10% of claims — compared to 32% in 2009.

Mr Haslam says many customers with fraudulent intentions simply do not call back after being informed their claim has been flagged by VSA technology, and cites the software as an effective deterrent: "It identifies people not telling the truth, who either back out or go through the full process. For a fraudster, to hear that their voice is going to be analysed is quite a tough message to take." He insists the use of VSA, with its incurred cost, has proved good value for money: "The cost has been paid back in terms of increased fraud detection, reduced complaints and a better customer journey, several times over. There is no doubt that this has been a massive success for us."

Another high profile name to back the technology is Highway. Back in 2002, the insurer was one of the first companies to use the software, after a successful trial. Following its acquisition by LV in 2008, the insurer has continued to use the software, and LV has been so impressed by VSA that it rolled out the software to its LV Direct offering in March 2009.

Ursula Coulibaly, head of group financial crime at LV, says the technology has surprised her — and insists it is a vital asset in the enlarged insurer's fight against fraud in fire and theft claims. Ms Coulibaly adds that Digilog has sped up the claims process by 2% to 3%: "We make sure VSA is not used in a negative way; what we have been able to do is initially pick up on suspicious fraud and speed up the process for innocent customers. Because we know there are smaller risks associated with certain claims, we are able to process these more quickly. It may be that in the future we are able to look at other product lines and parts of motor claims."

However, after Digilog's initial explosion onto the market five years ago, several high profile insurance companies have withdrawn support for the technology — including RSA, Direct Line and Halifax Bank of Scotland General Insurance — with its validity now questioned by many in the industry.

Phased out involvement

HBOS GI confirmed in April 2004 it was to use VSA software on its home insurance claims, after conducting a trial to determine the effectiveness of the software. The pilot conducted by the insurer found that its use identified 30% of household claims were worthy of further review, with 40% of those claims subsequently withdrawn by the applicant. Today, HBOS GI, operating under Lloyds Banking Group, no longer uses this software in its claims handling process, and phased out its involvement with Digilog by the end of 2007.

Dave Berry, manager for financial crime prevention at Lloyds Banking Group, says the decision to distance itself from VSA software was to provide a more "human" approach to its customers: "It didn't give that same level of rapport as talking to someone in a normal conversation management style approach, and we moved more towards cognitive interviewing. The issue is that voice stress makes you reliant on the technology and, because you can't guarantee 100% accuracy, it is better from the customer's perspective to have a more friendly approach," he adds.

In response, Provident's Mr Haslam insists that Digilog's product does not rely on technology alone, and works in tandem with cognitive questions posed by staff. "It is very much a two-pronged approach. The VSA may pick up some risks, but that is very much backed up by the human who has been trained and licensed to ask these cognitive questions. When you put the two together, they can be quite a powerful tool; a human can have an off day, whereas a voice risk analyser can't."

Mr Berry reveals other factors also lay behind the decision to cease its use. Cost was a major reason, he explains: "Cost will always be a factor. I have to apply budgets proportionately and appropriately, and I believe that the most appropriate way to spend money for HBOS and Lloyds is on developing our colleagues." He also says the insurer was keen to not treat all of its customers as 'potential criminals'. "It is about the proportionality of investigation. You don't investigate every single claim that comes into the operation, and don't treat every customer as a potential criminal, far from it."

Mr Berry says the success of HBOS's VSA-based system has now been matched — and bettered — by its new strategy, adding that this type of technology must progress in order to convince sceptical insurers: "It has to demonstrate that it has moved on from where it was before. It has to demonstrate that it is customer focused, and it must prove itself a sustainable, reliable proposition that identifies fraudulent claims rather than flagging up the wrong type of risk."

Others have been more open in their criticism of VSA software, citing its lack of scientific credibility. Several academics have incurred the wrath of Nemesysco — the provider of Digilog's voice analysis engine — with some studies in the US even suggesting Nemesysco's technology operates on nothing more than a 'chance probability' of accuracy.

One such critic is Professor Anders Eriksson, a professor of phonetics at the University of Gothenburg. Professor Eriksson published a paper that strongly challenged the validity of Nemesysco's voice risk analysis product. Nemesysco claimed the paper was libellous, and the publication was subsequently removed.

Scientific scepticism

But Professor Eriksson still maintains his views — that the technology is not fit for use in a multi-million pound industry. He tells Post: "There is no validity at all in the technology itself. It is no better than flipping a coin. The lie detector itself does nothing. We took the technology apart in principle and tested it. There is no principle at all attached to anything scientifically known. It doesn't make any distinction between you or a dog or a bus. Companies would save or earn more money if they invested the same amount in training their own people at being better at judging these claims."

One of the criticisms cited by Professor Eriksson is the 'bogus pipeline effect', which attempts to explain VSA's success by noting customers are more inclined to tell the truth if they know they will be monitored. This criticism, therefore, points to the technology as a placebo. He adds: "People believe it works. So they are more reluctant to lie if they think they are being monitored by a 'real' working lie detector."

Under UK law, insurers must notify customers if a call is being monitored. And Provident and LV both reveal they hint to customers VSA is being used on a phone call.

Bobby Gracey, vice-president of global counter fraud solutions at loss adjuster Crawford & Company, has recently reviewed the anti-fraud techniques used by insurance companies, and warns those overly reliant on such software. He suggests the successes notched up by using it do not necessarily point to flawless and accurate VSA technology: "I think the jury would be out on what causes Digilog's success. Is it that cognitive element of the script, or is it the technology? Academic papers would suggest it has nothing to do with the technology. I would be interested to hear the views of Nemesysco on the academic merits they could share that would respond to these findings."

Kerry Furber, managing director at Digilog, says the bogus pipeline effect is not a valid explanation for his firm's successes: "Companies that aren't using our process have the same requirements to tell a customer that they are monitoring a call. Those are basic Data Protection Act requirements. I don't think the warning has an impact on customers. If you look at the results our clients are achieving, they are returning 30-to-one on their investment. Those that criticise the technology have their own vested interests."

Mr Furber refutes allegations that VSA technology is flawed, and insists criticisms from some academic studies are short-sighted: "Those papers only concentrate on the technology on its own. We run Nemesysco's product alongside something we have developed, called 'narrative integrity analysis', which looks at verbal signposts including problems in answers that customers might give.

"In our opinion, the technology has to work in controlled realms, as we need to ensure that the questions asked of individuals are appropriate, and it would be wrong to cause them stress," he adds. Mr Haslam agrees, reiterating that Digilog's product does not totally rely on VSA alone, and works in tandem with cognitive questions posed by staff.

Furthermore, Mr Furber claims Digilog is currently in negotiations that, if successful, would double its number of insurance clients, which now stands at 15 companies. Interestingly, he also claims to have seen a significant surge of interest in the technology over the past few months.

Post understands a number of specialist over-50s insurers are interested in implementing VSA in the claims application process, and Admiral is thought to be trialling VSA software after initially becoming involved with Digilog in 2003.

Mr Furber refuses to reveal the identity of any firms interested but he attributes the surge of interest to the challenging economic conditions that usually precipitate fraudulent activity: "For far too long the insurance industry has concentrated too much on trying to detect premeditated fraud, and has failed to understand that the bulk of fraud is opportunist," he claims.

The issue of voice stress analysis, and Digilog's presence in the insurance market, is clearly divisive. The insurance industry, so shrewd at monitoring risks, is clearly tentative about taking a gamble on the technology, at least at the moment.

However, as the effects of the economic downturn are expected to see fraud rise further, the industry may turn to emerging technologies to stem the tide. It is evident that many insurance companies still look at their staff as the most important line of defence. As Mr Gracey puts it: "If you educate people and commit budget and resources to education programmes and institutionalising fraud strategies within organisations, then you stand a good chance of defeating fraud. People are always your greatest asset."

The debate over the validity and effectiveness of VSA is therefore ongoing, and only time will tell who will profit from embracing it.

  • LinkedIn  
  • Save this article
  • Print this page  

You need to sign in to use this feature. If you don’t have an Insurance Post account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: