Merchants' fraud filters are rejecting more legitimate orders, Adyen finds

Most businesses measure fraud through losses, chargebacks and security incidents, according to Mark Beare, head of consumer at Malwarebytes, writing in Forbes ^[9][10]. He wrote that they are much less equipped to count the customers and revenue lost when people fear fraud, or when the systems meant to protect them get in the way ^[9]. An abandoned purchase leaves no chargeback, fraud alert or incident ticket behind ^[16]. A fraud team can lower its chargeback rate by tightening filters, and its dashboard will score the result as progress.

Both sides of that trade-off cost money, and only one side gets booked. If Riskified's three-to-one ratio is even roughly right, the unbooked side is the larger ^[4]. The Datos Insights forecast says it is also growing: the move from $231 billion to nearly $265 billion is a rise of about $34 billion, roughly 15 percent, in one year ^[5][6]. Adyen's report put the choice plainly. "The question is no longer how much fraud a business is willing to tolerate," the study concluded. "The question is how much legitimate revenue it's willing to lose trying to stop it." ^[8]

A skeptic would ask who is supplying these numbers, and the question is fair. The column's author is a Malwarebytes executive, and he cites his company's own June 2026 research on AI-generated product photos and reviews ^[10][17]. The ratio comes from a single fraud protection firm. The $231 billion figure is a forecast made in May 2024 ^[5]. As quoted in the column, Adyen's evidence on declines is the share of merchants who say they rose; it does not put a dollar value on the orders lost ^[2]. Still, the direction holds across sources with different interests. LSEG Risk Intelligence's March 2026 survey of more than 21,000 adults found that 97 percent of responding fraud victims changed their behavior after being scammed ^[11].

The incentive problem sits in the org chart. The fraud team owns the chargeback rate. The lost sale, if anyone sees it, turns up as abandonment in a commercial team's funnel. Beare wrote that companies "can and should start measuring fear and trust with the same rigor as revenue and begin treating them as quantifiable business variables rather than soft concerns." ^[12] The decision available this quarter is modest: report false declines and trust-related abandonment beside chargebacks, under one owner. The consequence comes a quarter later, when filter thresholds become a number that risk and revenue leaders negotiate and the fraud team is graded on legitimate sales kept as well as losses avoided.

AI adds a second source of hesitation. A 2026 Usercentrics study of 11,000 customers in seven markets found that close to half had taken at least one action with a direct financial consequence for a brand in the previous six months because of concerns about how AI was using their data ^[13]. The same study measured stated willingness to pay: 52 percent said they would pay more to brands that are transparent about AI and their data, at an average premium of 7 percent, and among 18-to-29-year-olds the share rose to 67 percent ^[14].