Callcredit taps machine learning from Featurespace to fight fraud

Callcredit, the UK's second largest credit ratings agency, has selected technology from Cambridge-based start-up Featurespace to help crack down on identity fraud.

Featurespace, which was co-founded by Cambridge mathematics professor Dr Bill Fitzgerald and his former PhD student David Excell, uses the statistical technique of "Bayesian inference" to make predictions from data.

Rather than extrapolating a trend from incomplete data, Bayesian inference starts by assuming that the likelihood of any outcome is random, but refines its estimate as new data becomes available.

"The Bayesian approach tells you the probability that a hypothesis is correct when you can’t do random trials or reproducible experiments,” Fitzgerald told Information Age last year.

Callcredit offers its customers fraud detection, among other services. One of the ways it does this is to analyse customer buying patterns to spot unusual and potentially fraudulent behaviour.

According to John Cannon, head of ID and fraud at Callcredit, the company tested Featurespace's ARIC product on historical data and found that it was much more effective than its existing analytics.

"They were able to find the fraud very effectively, and it means you end with a lot fewer false positives," he said

Callcredit has therefore decided to deploy ARIC within its fraud analysis unit. It also plans to use the system to help prevent fraud in its consumer wing.

Callcredit runs a service called Noddle, which allows consumers to access their own credit reports. Credit reports contain data that could be lucrative to fraudsters, though, and the company will use ARIC to beef up the process of identifying and authenticating customers. 

The two companies also plan to offer a joint service, which allow commercial clients to apply ARIC to Callcredit's credit rating data. This will allow them to predict the likelihood that a customer will default on a loan, for example.

Commercial clout

Featurespace was founded in 2005, and Cannon says the companies have been in discussions for a number of years. Last year, though, Featurespace received fresh investment from Invoke Capital, the VC firm set up by former Autonomy CEO (and another former student of Fitzgerald's) Dr Mike Lynch.

In October 2012, it appointed Martina King as its new CEO. King was previously managing director of Aurasma, Autonomy's augmented reality spin-out, having previously managed Yahoo!'s UK division and London radio station Capital FM. Four of the company's five commercial employees are former Autonomy workers.

According to Cannon, King's involvement in the company has made it more commercially viable proposition.

"Martina has brought a heck of a lot of credibility," he says, "The clever software, the clever analytics, all of that is fantastic. But for a partnership to work, the people are equally important. And these are people we want to do business with.

"We think Featurespace will be very successful."

It remains to be seen whether Featurespace's link to Lynch is a boon or a burden. In June, financial website quoted unnamed sources as saying that the FBI's investigation into Hewlett-Packard's accusations against Lynch and the former Autonomy management "will drag on for a couple of years".

Last week, Featurespace's first customer, betting exchange Betfair, signed a five-year agreement with the company to continue its use of the technology.

Pete Swabey

Pete Swabey

Pete was Editor of Information Age and head of technology research for Vitesse Media plc from 2005 to 2013, before moving on to be Senior Editor and then Editorial Director at The Economist Intelligence...

Related Topics