It’s early in 2017, but fraud looms like Nosferatu chancing upon the sleeping maiden. And lest the queasiness factor of chargebacks and outright damage to merchant reputations have executives reaching for the Maalox — or something no less liquidly but a lot harder — a few sobering stats are here to keep us cold-eyed as we journey into the latest Data Drivers, between PYMNTS’ Karen Webster and Jason Tan, CEO and cofounder of Sift Science.
Yes, that is billions with a B, and a capital B, as it represents the staggering amount that merchants lost last year, in the United States alone, from false declines, as estimated by BI Intelligence. This is the amount that went by the wayside from legit customers because merchants thought they looked shady, as Webster noted.
Looking at fraud systems, said Tan, the impetus is to err on the side of “minimizing chargeback and payment fraud,” but traditional systems, he said, “don’t do enough to manage the tradeoff of making sure there is an acceptable and healthy rate of false [declines].” He likened this approach to casting a very wide net, at the expense of letting innocent people accomplish what they want during a transaction. As an analogy, he said, think of an airport security setup, where the clear majority of users are not terrorists, yet the current system demands an unpleasant and inefficient process.
As so much business moves online, he said, with that $8 billion figure (writ much larger internationally), the question becomes how to minimize threats while maximizing revenues without forcing the tradeoff of security and convenience for the customer. The ability to provide convenience will become “a huge competitive differentiator” over time, he said, citing Amazon, with its famed quick checkout process, in turn predicated on the ability to manage risk accurately and in real time.
One salve: Machine learning can help these businesses operate smoothly in making choices about who they can and cannot trust.
86.1%: Percentage of online merchants who expect to see the same or more fraud and abuse attempts in 2017.
This data point begs the question: What does the remaining 14 percent know that others do not? That’s only a tongue-in-cheek statement, as proffered by Webster. Said Tan, this is his firm’s own data, and even adjusting for sample bias, that 14 percent depends on the business, with reliance in part on systems that are already in place. There are many other firms that do not count chargebacks as their main problems with fraud, instead, he noted, working against account takeovers. But of course, the large preponderance of fraud points to real concern. As Tan said, as the world moves from offline to online, “every crime that happens in the real world is going to happen online. We must prepare for that inevitability. The power of the online world is that you can collect and put data to use that you might not have available offline … Having large quantities of data can really drive improvement in terms of decision-making.”
An intelligent system can scale and keep up with fraud. Think of email spam filters of old that were rules-based, he said, and inefficient, as those rules were rigid, poised to capture, for example, emails with the word “Viagra” in them. Now, he said, more sophisticated systems, such as Gmail, can work with rules and change them at the same time to analyze statistics and probability so that, as spammer behavior changes, the software can step up and stay ahead of the malefactors. The same scenarios can extend to fraudulent behavior and transactions, “a harmonious hybrid of humans and machines … The machines are not perfect,” and so, machines must work with humans, relying on the latter to tell them what is right and wrong.
48%: Percentage of online merchants who observed a rise in Account Takeover (ATO) last year.
This comes from Sift Science, and Tan said he would have expected the number to be higher, going so far as to say that account takeovers would be the future of fraud. He likened account takeovers to the perfect crime, as the identity being used is one that is a strong one, trusted by the merchant. The fraudster parachutes in, and with the positive reputation as a lure, data becomes tempting to steal, too. So, again, machine learning is a boon, he said, cautioning merchants to “leave no stone unturned … even the most inconsequential piece of data absolutely means something in the grander scheme of things.” And it is the machine that will help a user figure out what matters and what doesn’t — as the machine flags what may be subtle and worthy of examination, while it is the human who looks.
For while, he said, a hacker needs to find only one way into a system, it is incumbent on the merchant to plug all the holes. “Don’t fear technology,” he said, “as it can be the best supplement to your business.”