One of the UK’s biggest credit reference agencies has revealed it is using Microsoft’s artificial intelligence to spot fraudulent loan applications and protect consumers.
Callcredit, which works with lenders across the UK, is using Azure Machine Learning to identify criminals who pretend to be other people when they try to access credit reports and borrow money.
The service has been successful in stopping fraudulent access to Callcredit’s credit reporting and scoring service, named Noddle, and protecting consumers from subsequently having bad loans taken out in their name.
The Bank of England revealed that Britons added £1.5bn to their debt in October, the latest available figures, taking the total to more than £205bn.
“We were using a third-party, specialist machine-learning fraud-detection platform to help us spot fraudulent sign-ups to Noddle,” said Mark Davison, Callcredit’s Chief Data Officer. “We had to decide whether to bring this in-house; so, we created a model internally, on Azure Machine Learning, which outperformed the third-party product. We’ve now retired the third-party service and are running our own machine learning-based fraud detection, spotting how fraudsters are trying to sign up to Noddle and preventing them getting access to those credit reports.”
Noddle is a free for life credit report and score service that searches the market to find credit cards and loans available to a person based on their credit rating, and gives them a view of the likelihood of being accepted for them. Callcredit is using Azure Machine Learning, which is part of Microsoft’s cloud service and uses existing data to spot patterns and predict future events, to identify if people signing up to Noddle are who they say they are.
It has also freed up Callcredit’s fraud detection specialists to focus on more sophisticated patterns of criminal activity as well as provided a proof of concept to enable the company to offer the service to its wider client base.
The second largest credit reference agency in the UK has also used Azure Machine Learning to spot people who will fail to pay back loans - either intentionally or because of legitimate financial problems that occur after taking out the loan. A trial that was run by the company using Microsoft’s technology cut the level of default in a portfolio of 60,000 credit cards by more than £1 million because it was better at spotting people who would default on loans, Davison added.
“It helps lenders make more accurate decisions, more quickly. The loan applications a lender will turn down and the ones they will approve are easy to spot. The grey area in the middle, which are the ones they have to manually refer, are much trickier. As a result, the end consumer goes through a far more onerous process where an underwriter has to review the case. That’s not a great experience for the consumer or the lender. When you decide you need a credit card or a secured loan for a car, for example, you want a decision quickly. Being able to give more accurate predictions to lenders enables them to make that decision really quickly.
“People don’t realise how much our service is empowering their daily lives and making it much easier to access the right financial products. They see the car, they apply for the loan and whomever is lending to them needs to make a decision really quickly. That real-time, instant decision has to come from well-managed data wrapped with very sophisticated insight and analytics.”
If lenders are able to be more accurate when predicting who will default on loans, “that has a beneficial effect for the economy, too”, Davison added.
Yorkshire-based Callcredit was founded by Skipton Building Society in 2000 and is owned by private equity firm GTCR. It employs more than 1,300 people in Leeds, London, Stockport, Warrington and Kent, and has international offices in Spain, the US, Dubai and Lithuania.
Contributed | what does this mean?