Fraud Detection & Prevention

Featurespace: 'Accidental' revenue via machine-learning-driven fraud detection

One of the most frustrating holiday experiences I ever had was when my bank decided to lock me out of my account the minute I touched down at my destination. In short, it decided my financial activity was suspicious, my attempt to withdraw cash in Italy was probably fraudulent, and took a whole load of expensive phone calls to convince it otherwise.

This is precisely the kind of situation that Featurespace is helping organisations to limit. “You can’t stop 100% false positives,” – i.e., these untrue notifications of fraud – says the CEO Martina King, over the phone from the company headquarters in Cambridge, UK. “But we lead to a 70% reduction.”

Featurespace was co-founded by the late Professor Bill Fitzgerald who the Independent described in his obituary as: “The Influential Cambridge professor who did ground-breaking work in the field of signal processing”. The solution is grounded in machine learning and based on the concept that via data, maths and computer modelling, you can learn what “normal” is in order to flag genuinely abnormal activity.

This makes it a kind of sister to what Darktrace does in the cybersecurity field. And Dr Mike Lynch – who is director at Darktrace – was taught, in fact, by Professor Bill Fitzgerald.

“It was the gaming sector [Betfair was its first client] that took us out of the universities,” says King. While other clients include KPMG, TSYS, which processes MasterCard and Visa transactions, along with addition of “a major US bank” which remains undisclosed.

“Any data that is relevant can be fed in,” says King, former Managing Director of Augmented Reality company Aurasma – which was founded Dr Mike Lynch – and she left once it was acquired by HP. In a career that has spanned Yahoo and Capital Radio, King was named one of Silicon Republic’s 40 most powerful women leading tech earlier this year.

“The thing that has surprised me,” she says “is that I thought we’d be a fraud detection company. But we’ve become a revenue generating company.”

This hinges on the high cost of getting things wrong. If a company doesn’t really know what is happening with its data, but flags suspicious behaviour based on, say two predetermined rules, it came only prove correct some of the time.

“Far more genuine customers get caught up in the general rules,” than get impacted by fraud, explains King. So if you reduce inaccuracy by better data modelling the business will become more streamlined and as a side effect more cost efficient.    

Featurespace has raised $16.73 million in funding since 2012 including $9 million this Spring. This was led by US fintech investor TTV Capital – and the company intends to use the money to open a US office. “We have been working with US customers before opening an office,” says King.

Machine learning is gradually getting used in lots of different industries and areas but still often gets muddled up with the wider concept of AI and the rise of robots. Yet King does not see this as a problem when explaining the benefits to clients. If you talk about data and the business uplift, she says, “the how becomes less important”.

“The insurance market is beginning to become susceptible to machine learning,” adds King and there is still an “awful lot to do in the financial world”.

“It is exciting because it is still so pioneering,” she says. “And it is all home-grown technology.”  

I can’t help wondering, if despite differences, King feels part of the wave of AI startups coming out of the UK at present. She says the analogy she likes to use is Arm:

“Arm is ambitious, it’s from Cambridge and without it mobile phones wouldn’t work,” she concludes.


Also read:

Why is the UK such a centre for AI innovation?

Can ‘good’ machine learning take on global cybercrime?


« C-suite career advice: Rudi de Sousa, YLD


Database architects help scientists deal with data floods »