Today’s fraudster is as clever as ever. In fact, as commerce continues to shift online, bad actors are seemingly always a step ahead. They’re gaming the system and causing headaches for businesses—especially those with card-not-present (CNP) and online transactions.
As high-grade security through EMV chip card technology has been adopted in stores and our shopping habits have moved online, so has fraud. The shift has been incredibly fast.
eCommerce sales in the United States represent 13.5% of all retail sales and is growing at an annual rate of nearly 9%. With that growth, merchants become potentially liable for more fraudulent activity. In fact, worldwide card fraud was $28.8 billion in 2018 and forecasted to grow to $42.3 billion by 2026.  The majority of these losses are related to CNP transactions which typically are the merchant’s liability.
As merchants evolve to meet customer demands for a pleasant online and omni-commerce experience, they must consider how to protect themselves and their customers. The rapid adoption of new technology has sometimes outpaced the ability to make new platforms secure. For instance, a restaurant’s new online ordering app may have fraud vulnerabilities but the opportunity cost of not having a mobile platform outweighs that risk.
A merchant may believe he or she is taking all the right steps to protect their business, only to be frustrated by lackluster results. That’s because fraud attacks are so dynamic, traditional tools generally aren’t advanced enough to stop the fraudulent transactions while ensuring valid transactions are approved.
Many merchants in this uphill battle seek help from their payments processing partner. Working with a payments fraud expert can help alleviate the frustrating cycle of fraud loss that a retailer can experience.
In a fraud loss cycle when losses are growing, a merchant often institutes more aggressive fraud rules to reduce the financial impact—erring on the side of detection and protection versus permission. But these rules can backfire. They will reduce some instances of fraud, however, they are also highly likely to decline good transactions.
Because traditional fraud prevention tools use a rule-based mechanism for action, they aren’t nimbly evaluating each transaction. Rather, they’re seeing if the transaction checks off a certain number of criteria to make decisions. For instance, the country where the purchase is coming from or the total amount purchased in a certain span of time can be red flags. With such a rigid structure in place, clever fraudsters can find loopholes and workarounds while innocent people can suffer if their purchase transaction anomalously checks off too many of the boxes. Further, this results in false-positive declines where good transactions get flagged as fraudulent.
How big is the impact of these false positives? A recent study by Javelin Strategy & Research and Vesta Corp. found that 30 percent of declined transactions due to suspected fraud are believed to be authentic. All of those false-positive declines don’t only cut into potential revenues, they can erode consumer trust in the businesses and their systems. In fact, a study by Oracle found that 89 percent of consumers turn to a competitor after a negative or poor experience with a business.
From losing out on repeat customers to seeing a decline in customer loyalty, false-positive declines can undercut the steps a business owner takes to reduce instances of fraud. If the merchant loosens the fraud rules, he or she may be liable for more chargebacks. However, if the retailer filters too aggressively, he or she may unintentionally punish customers and eat into sales. It’s a Catch-22, but there is a way to successfully address this.
Moving away from rules-based fraud protection, savvy merchants are embracing machine learning to stop the fraud loss cycle. Because machine learning is designed to automatically respond to variances in data, behaviors, trends and more, it’s an agile solution. It relies on the integrity of the data source it’s built on. Given a large pool of transactional data, a machine learning platform is engineered to build predictive behavior models and adjust on-the-fly for new fraud trends. Its system objective is to understand the patterns of fraudulent and legitimate transactions because of that data. And every new transaction becomes a potentially new learning event for the machine learning model.
The idea behind a machine learning platform is that it makes it possible for merchants to not only push back against the rising fraud tide, but to ensure their systems are optimized so good transactions are approved without a hitch.
Worldpay learned the benefits of machine learning several years ago as it moved from a rules-centric system to a machine learning-based fraud system. The prior system consisted of over 300,000 rules to protect its customers and the overall performance was middle-of-the-road. Not to mention the amount of effort to maintain such a large rule set. With a focus on high quality machine learning, Worldpay was able to reduce its rule set to less than 3,000 and has become an industry leader.
By adopting a machine learning fraud system, businesses can better help their customers enjoy the seamless, convenient experiences of new commerce modalities like mobile or digital assistant purchasing. At the same time, business owners can have peace of mind knowing they’re increasing profits and finally breaking the cycle of fraud loss.
 eCommerce report from Statista, 2018
 Nilson Report, Nov 2018
 The Financial Impact of Fraud: Merchants Challenged as E-commerce Fraud Rises Post-EMV; Javelin Strategy & Research and Vesta Corp.; October 2016
 Customer Experience Impact Report, Oracle