Your fraud rules are quietly rejecting paying customers

Fifty wrongly declined orders at a 100 euro basket and 30 euro of margin wipes out 1,500 euro. Stop 1,000 euro of actual fraud in the same period and the programme is already underwater before anyone has paid the licence fee for the tool doing the stopping. That arithmetic, set out in the Merchant Risk Council’s 2026 Global Ecommerce Payments & Fraud Report, is the clearest argument yet that a low fraud rate is not evidence of a healthy checkout. It can just as easily be evidence of a checkout that turns away people who wanted to buy.

The survey behind it is substantial: 1,278 responses from 37 countries, collected in December 2025. The headline shift is in what respondents say they are optimising for. Twenty-nine per cent now name minimising operational cost as their top fraud-management priority, against ten per cent two years ago. Read that cynically and it is budget pressure. Read it properly and it is a maturing of the discipline: nearly one in three teams has stopped treating a fraud number as a score to drive towards zero and started treating it as one line in a profit calculation that also contains declined good orders, manual review hours and abandoned baskets.

Where the cost actually lands in the funnel

The expensive part of fraud prevention is rarely the subscription. It is what the rules do between the payment button and the confirmation page. An over-tight rule set produces three distinct kinds of damage, and only one of them shows up in a fraud dashboard.

The first is the hard decline. The order is refused, the shopper sees a generic error, and in most webshops that error says nothing useful. There is no message explaining what to try next, no alternative payment method surfaced, no second attempt path. The shopper reads it as “your card does not work here” and goes to a competitor. That buyer is not recorded as a lost sale anywhere in the analytics stack, because the session simply ends.

The second is step-up friction applied to people who did not need it. Extra verification, an identity check, a 3-D Secure challenge triggered on a repeat customer with three years of clean history. Every added screen in the payment phase costs completion, and the drop-off is sharpest on mobile, where a redirect to a banking app or an authentication screen can break the session entirely if the return journey is slow or the app has to be fetched from the background. The shopper who abandons at an unnecessary challenge is counted as cart abandonment, not as a fraud cost, so the fraud team never sees the bill.

The third is the delay. An order held for manual review is an order that does not ship today. The MRC puts the average cost of resolving a first-party dispute, where a genuine cardholder disputes a transaction they authorised, at more than 80 dollars. Manual review carries its own hourly cost, and when reviews pile up at peak the queue itself becomes the friction: confirmation emails go out late, delivery promises slip, and the post-purchase experience that drives repeat business starts in the wrong place.

Card testing is a checkout problem, not just a security one

The report’s note on card testing deserves more attention than it usually gets in a conversion discussion. Attackers submit large volumes of stolen credentials to find which ones still work. They are not trying to buy anything. But the damage is felt by real shoppers: payment pages slow down under the load, authorisation rates drop, and the acquirer starts paying close attention to the account. A payment form with no rate limiting, no bot detection on the card field and no minimum-value guard is effectively a free credential-testing service that your paying customers subsidise with a slower, flakier checkout.

The practical fix is placed before the payment step rather than inside it. Rate limiting per IP and per device, blocking repeated low-value attempts, and keeping gift-card balance checks and zero-value authorisations behind some form of challenge removes most of the volume without touching the path a real buyer walks. That is the pattern worth generalising: friction belongs on the signals that look like automation, not on the signals that look like a human with an unusual delivery address.

Measure the decline, not just the chargeback

Most merchants cannot answer the question the MRC data implies, because the two halves of the number sit in different systems. The chargeback rate lives with the fraud or finance team. The declined-order rate lives with the PSP reporting. The abandonment rate lives in analytics. Nobody owns the sum.

Three things are worth doing this quarter. Pull the declined and refused transactions for the last ninety days and sample them manually: how many look like ordinary customers with an ordinary basket? Every one of those is lost margin that your fraud rate will never reveal. Second, instrument the step-up. Log how many shoppers are sent into additional verification, and how many of those complete. If completion after a challenge sits meaningfully below completion without one, you have priced the challenge and can decide whether the risk it catches is worth it. Third, build the exemption side deliberately. Low-risk transactions, returning customers with stored credentials, and recurring purchases can often skip the challenge entirely, and a frictionless flow for your best customers is cheaper than any rule that treats them like strangers.

The shift the MRC picked up is really a shift in who the fraud programme serves. Tuned for a zero fraud rate, it serves the risk report. Tuned for net margin, it has to account for every good order it refused, every basket it scared off and every review queue it filled. That second version is harder to build and considerably more profitable to run.

Source: practicalecommerce.com

Related reading