Machine learning is helping merchants and financial institutions evolve in the fight against online payments fraud. Here’s a look at the current state of payments fraud and how to fight it effectively.
Payments fraud remains a substantial problem for all parties in the ecosystem. According to Experian’s 2017 E-commerce Fraud report, online shopping fraud attacks increased 30% in 2017 vs. 2016. The heart of these attacks is often a data breach; in 2019, we have already seen more than 50 confirmed data breaches of well-known companies.
As records and PII are stolen and listed for sale on the dark web, ecommerce fraud attacks spike. Consumers, merchants, and banks alike all feel the ramifications of breaches and fraud. Consumers may have their identities stolen; merchants lose revenue to chargebacks and false positives, and banks lose consumer trust and incur costs associated with dispute resolution.
As m-commerce continues to grow, mobile fraud will, too. According to the LexisNexis True Cost of Fraud study, mid-to-large m-commerce merchants that sell digital goods have higher fraud-related costs than other merchants. For these merchants, every $1 of fraud costs an average of $3.29—almost a quarter (24%) increase over 2017.
Despite the sharp increases in online payments fraud, ecommerce continues to be a preferred channel for purchasing. Advancements in machine learning (ML) have also offered promising solutions to the fraud problem.
ML enables the fraud prevention process to be streamlined with the help of automation. With ML businesses can employ analytic techniques that “learn” patterns without the help of a human. Through artificial intelligence (AI), these analytics capabilities can be applied to fraud prevention through the identification of fraudulent transactions.
Just last month, Stripe announced the launch of its new machine-learning based fraud protection tool for chargebacks. Radar prevents fraud and also automatically reimburses merchants for dispute fees and charges. Traditionally, merchants are tasked with finding and presenting evidence to fight disputes. Evidence collection is a time-intensive process that can cost more than the original dispute amount itself.
ML can learn transaction patterns and begin to determine, very effectively, which transactions are more likely to be fraudulent. Not only does this prevent more fraud, but it also prevents and reduces false positives. Considering the large volumes of transactions that many businesses process, it’s easy to see how automated detection can eliminate some of the costs associated with fraud and fraud prevention. An additional benefit is that ML models continue to learn and adapt, making them adept at identifying new fraud tactics. Overall, ML enables:
Real-time decision-making — Unlike traditional rule-based methods, ML does not rely on ad hoc rules to determine what orders to reject or accept. This eliminates the manual interaction needed and can facilitate the evaluation of large volumes of transactions in real time.
Fast response to change — ML is constantly “learning” when analyzing and processing new data, so it can self-update to adapt to new fraud tactics and techniques.
High accuracy — ML is typically more effective than humans in identifying subtle fraud patterns. This is especially important as cybercriminals grow increasingly sophisticated and stealth. ML can also use its ability to mitigate false positives.
Cost reduction—ML is highly scalable and the technology costs associated with ML have gone down over time. By improving fraud detection accuracy and minimizing false positives, Ml can significantly reduce the costs associated with online fraud.
Perhaps most importantly, ML is making strides in keeping payments frictionless for consumers. Legacy systems and fraud prevention tools have been notorious for interrupting payments flow, or, at the very least, presenting a minor nuisance to consumers. Employing machine learning as part of a multi-layered fraud prevention solution improves capabilities without slowing down the payments experience.
This is essential as Experian reports that 35% of prospective customers leave online applications because of friction. While a fraction of those may be fraudsters, the greater percentage will be legitimate customers looking to make a purchase. ML and other frictionless tools can easily and seamlessly identify fraud trends without bothering customers. The identification of patterns happens behind the scenes and allows merchants to fine-tune fraud prevention strategies while also learning more about consumer behavior patterns and preferences.
Online payments fraud must be combated with technologies that are intelligent enough to keep up with the dynamic fraud landscape. The key is to strike a balance between the right mix of fraud tools without inhibiting or adding friction to the customer experience. Tapping into ML can streamline the fraud prevention process by removing certain human components, analyzing large volumes of transaction data, and automating pattern recognition. The result is better accuracy, real-time fraud prevention, and cost reduction.
Implementing frictionless fraud prevention tools and strategies can significantly impact the bottom line. Both merchants and financial institutions must look to next-generation tools for better validation and authentication. ML is one piece of the fraud prevention puzzle that will enable organizations to evolve alongside the fraud landscape and stay one step ahead of cybercriminals.