Led by Amazon in the US, which accounted for nearly half of all of American ecommerce sales in 2017, online sales are a growing part of consumers’ buying habits, worldwide. These changing habits also present increased opportunities for fraudsters, who use evolving tactics and stolen payment card information to commit e-commerce fraud. FICO’s European Fraud Map shows that card-not-present (CNP) fraud, which includes ecommerce, represents some 70% of the card fraud in Europe, and we see similar high percentage rates seen in many parts of the world.
The risk of CNP fraud, like e-commerce sales, is going in only one direction: up. According to data from Statista, ecommerce sales will reach $4.5 trillion in 2021, from around $2.3 trillion in 2017.
Our new CNP models for FICO® Falcon® Fraud Manager combat this trend with advanced machine learning (ML) algorithms. The new ML-enhanced fraud detection models show strong improvement in the real-time value detection rate, which measures the percentage of total fraud dollars detected by these models.
Machine Learning Finds More CNP Fraud Faster
The new CNP fraud detection models have been developed to detect fraudulent e-commerce fraud transactions as quickly as possible, to prevent significant fraud loss from an account. Organizations using Falcon will notice a much more aggressive Falcon score for the first high-value suspect transaction in a fraud episode, allowing earlier detection. Specifically, FICO’s simulation shows clients can expect an average score increase of 40% in the first high-value CNP fraud transaction using the new enhanced CNP machine learning models.
The new models have been quantifiably proven to improve the real-time value detection by at least 30% without increasing the genuine transaction decline rate. By building strategies that take into account the more aggressive Falcon score for the first high-value CNP fraud, Falcon clients can see a significant reduction in their fraud losses.
To view the chart click on this link
Courtesy of Scott Zoldi, FICO