$16bn annual savings using machine learning fraud prevention software
Oakhall, the London-based analysis firm, estimates that the cost of card fraud for the card industry increased 29% to $40.1bn in 2015 (2014: $31bn). This was primarily driven by the continuing rise in card use, a higher proportion of online shopping and the growing sophistication of fraud criminals. The study was published in conjunction with Featurespace, the global leader in machine learning adaptive behavioural analytics fraud prevention software. For the full study see the Cost of Card Fraud Report.
In 2015, costs associated with incidents of card fraud increased 34% to $21.8bn and fraud-related costs associated with genuine transactions declined, increased 24% to $18.3bn.
Jonathan Crossfield, Partner at Oakhall, said: “With card fraud costing the industry 7 cents for every $100 spent in 2015, we expect banks to continue to invest in more advanced fraud detection and prevention systems.
“80% of card fraud in the UK happens in “card not present” online or phone transactions where existing systems often trigger higher numbers of genuine transactions declined, resulting in customer dissatisfaction and lost income for the banks.”
Martina King, Featurespace CEO, commented: “Card fraud is becoming more prolific and far more costly - an almost $10bn increase in the cost of card fraud in just 12 months. In the UK alone, we will see card fraud continue to escalate, both in volume and sophistication.
“Our adaptive machine learning fraud prevention software has been proven to protect banks revenues and substantially cut operational costs from false fraud alerts. It also helps the banks maintain positive relationships with their customers.”
Using industry data, Oakhall estimates that global financial services firms could save at least $15.8bn (2014: $12.2 billion) annually by employing adaptive, machine learning fraud prevention software. The estimated savings comprising of $5.5bn (2014: $4.1bn) reduced fraud and $10.3bn (2014: $8.1bn) reduced fraud management costs and lost revenue.
Using historical customer data from a multinational retail bank, results showed that Featurespace’s machine learning software reduced undetected incidents of fraud by 25% compared to the bank’s existing fraud management system. The results also demonstrated that Featurespace’s fraud prevention system reduced the incidents of genuine transactions declined by over 70%.
By using machine learning, adaptive behavioural analytics software, card issuers can reduce genuine transactions declined, improve operational efficiencies and lower the incidence of undetected fraud, according to the Oakhall study.
Featurespace is a world leader in machine learning adaptive behavioural analytics software. Its services and products are employed in over 180 countries via a wide range of customers, including the leading US payments processor, TSYS, as well as Vocalink/Zapp, William Hill and Betfair.
London based Oakhall is a consultancy specialising in providing smart analysis and articulation services to private and public companies. Founded by previously top-rated equity research analysts, Oakhall has particular experience in the financial technology sector and focuses on bringing an analytical approach to market sizing and valuation.
Featurespace™ is the world-leader in Adaptive Behavioural Analytics and creator of the ARIC™ platform, a machine learning software platform developed out of the University of Cambridge.
Head quartered in Cambridge, UK, Featurespace has deployed ARIC to organisations that have services or products deployed in over 180 countries. Customers include Betfair, Vocalink/Zapp, Camelot, William Hill, and TSYS, the largest third-party processor of Visa® and MasterCard® credit cards in the U.S.
The ARIC™ engine – a real-time, machine learning software platform – monitors individual behaviours to catch new fraud attacks as they happen. The increased accuracy of understanding customer behaviour simultaneously reduces the number of genuine customers whose purchases and transactions are incorrectly declined.