Bitesize InsurTech: Featurespace
September 29, 2017 Chris Sandilands
Featurespace is a fraud detection software platform which analyses individual, real-time customer behaviour.
Founded in Cambridge University’s engineering department in 2008, Featurespace was created in response to a request by gaming company Betfair to outwit fraudsters by analysing online behaviour. It has expanded into industries prone to fraud, including insurance.
Featurespace logs customer behaviour on an insurer’s website as it happens and identifies suspicious actions, for example a user changing their age or profession after receiving a quote. This information is fed into a machine learning platform which uses adaptive behavioural analytics to calculate the probability of fraudulent activity and take actions, including alerts or outright application rejection.
The company tells us that eliminating fraudsters is the most effective way of reducing fraud (i.e. more effective than spotting fraud during a subsequent claim). They believe that they are particularly effective for aggregators, where people are shopping around most actively.
The company believes one of its key differentiators is that it analyses real-time behaviour (i.e., collections of actions over time) rather than single actions. This means that false positives are reduced while simultaneously blocking more fraud.
Featurespace insurance use cases include:
- Identification of fraudulent requests for quotes, mid-term alterations, claims or renewals
- Detection of ghost brokers
- Selection of customers that are less likely to claim, since customer behaviour on websites has been linked to likelihood to claim
Featurespace claims to reduce loss ratios between 2 and 12 percentage points.
The Oxbow Partners view
We recently hosted the Chief Data Officer of one of the large European insurance groups in London for a “data day”, where we introduced him to a number of selected data & analytics platforms. Featurespace was one of them.
We like the business and are impressed by its capabilities. Primary insurers who think they have a fraud problem should have a look at their proposition.