Fraud friction belongs on the moments an account changes hands, not on every payment
A fraud report can improve while the checkout gets worse for the people who actually want to buy. Sift’s figures from its global network show the payment fraud attack rate falling from 3.3% in the first quarter of 2025 to 2.8% in the first quarter of 2026. Over the same period the overall chargeback rate rose from 0.2% to 0.3%. Fewer attacks, more disputes. Sift’s explanation is patience. More attackers now invest time in an account before using it. For weeks or months it behaves like any ordinary customer and its purchases go through cleanly, until it is used for a single large fraudulent order. Its consumer research describes a second route into the same position, account takeover, where an attacker gains control of an existing customer’s account. Either way, the fraudulent order arrives from what looks like an established customer.
That pattern changes where friction does any good. The reflex after a bad month is to tighten checks at the payment step for everyone. An account with months of clean history, or one an attacker controls with its own login, passes those checks at the moment of payment as easily as any loyal customer does, because at that moment the two look alike. The shopper who feels the extra step is mostly the genuine customer, and that customer has a choice the attacker does not care about: leave.
Shoppers accept a check they understand and abandon one they do not
Sift’s consumer survey, run by Researchscape among 1,065 US adults in February 2026, contains two figures that look contradictory at first. 93% say they are willing to accept extra verification during login or checkout. Yet 73% say they have abandoned a purchase because of payment security concerns, and 52% say they would stop using a platform entirely after experiencing fraud.
Both can be true at once. Shoppers want protection and will accept a step that clearly provides it. What makes them leave is uncertainty: a page that does not explain what happens to their card details, a code request that arrives without a reason, a verification screen that fails and leaves them unsure whether they have paid. The difference sits in the wording and the timing far more than in the number of steps.
A code sent with a line such as “we sent this because you are signing in from a new device” reads as care. The same code at the payment step of a shopper who has ordered from you ten times, on the same phone, to the same address, reads as distrust or as a fault. Write every verification screen as if the person on the other side is a returning customer, because most of them are.
A false decline looks like a broken shop
A genuine shopper never sees a fraud rule. They see the outcome. Payment failed. An order confirmed by email and cancelled a few hours later. An account locked after a password reset. None of these carries the word fraud, so the shopper fills in the explanation themselves: a problem with their card, a problem with the shop, or a sign that the site is not safe. That last reading is exactly the security doubt that Sift’s survey links to abandoned purchases.
Three changes reduce the damage without telling anyone how your checks work. A declined payment gets a message that offers a next step, such as trying another method or contacting customer service, instead of a bare error. An order held for review tells the shopper it is being checked and when they will hear back, so the confirmation email is not followed by silence. An order you cancel after confirmation is reported to the shopper the same day, with the refund status, before they discover it on their bank statement.
The cost of a wrongly blocked customer does not appear in a fraud dashboard, which counts what was stopped and not who was turned away. Sift’s own benchmarking guidance makes the same point: a low fraud rate can be bought with too much friction, and a change in how often shoppers are stepped up changes the friction customers feel as well as the fraud outcome. Treat the share of shoppers who meet a check, and the share who complete after it, as checkout metrics, reported next to conversion and split between new and returning customers.
Move the checks to where the takeover happens
Sift’s report is explicit about what control of an account is worth. A card already saved, a loyalty balance waiting to be spent and the standing of a customer whose orders have always gone through can all be turned into money, and orders placed that way pass as genuine because the account itself is trusted. The weak point is therefore the account, and the moments when it changes: a login from a new device, a changed email address or password, a new delivery address, a newly added card, a loyalty balance spent in one go. Those are the moments where an extra confirmation is both justified and easy to explain to a genuine shopper.
The reverse also holds. A returning customer who reorders on a known device, to a known address, with a stored card, should meet nothing extra at the payment step. That is the journey stored credentials and one-click were built for, and it is the journey your best customers use most often. Keeping it smooth is not a concession to fraud, provided the account changes that typically precede an attack are the ones that get checked, and provided an unusual order from an established account, such as a first order far above its normal value, gets a second look after it is placed rather than a hurdle before it.
Start with an inventory. List every point in your account and checkout flow where a genuine shopper can meet a verification step, a decline, a hold or a cancellation. For each one, record last month’s numbers: how many shoppers reached it, how many completed, and how many contacted customer service about it. Rewrite every message that gives no reason or no next step, and move any check that applies to all payments to the account moments listed above.





