A checkout that works on the first attempt is worth more than a fast one
Every checkout depends on parts that can fail on their own: an address lookup, a delivery calculation, a fraud check, a single payment method, a payment page loaded from another domain. Most teams test the route on a day when all of them respond. The shopper meets it on the day one of them does not, and the order is decided by what the checkout does next. Whether that shopper retries, switches method or closes the tab depends far less on how fast the page loads than on whether the step still works.
That point came up at the National Retail Federation’s Big Show in New York in January 2024, in a session with Ludovic Maire, director of product management at the online fashion retailer SSENSE, and Curtis Crawford, global director and general manager of fintech and loyalty at Wayfair, as reported by Retail Dive. Maire’s point was that a shopper whose first attempt fails, or only loads slowly, may walk away. SSENSE handles up to 2,000 orders a minute during major product drops and can still process more than 99% of orders when one checkout component goes down.
The shopper only sees the outcome
Few webshops face a product drop of that size, but the planning principle applies to all of them. A shopper has no view of your infrastructure. They see a spinner that does not stop, an empty list where delivery options should be, or an error after they approved a payment. In Baymard Institute’s survey of US online shoppers, 17% say they abandoned an order because the website had errors or crashed, the same share that blames a checkout that was too long or complicated. A broken step costs as many orders as a clumsy one, and it is usually cheaper to prevent.
Design the checkout for the part that is down
Resilience sounds like an engineering subject, but most of what a shopper meets when something fails is design. Decide in advance what each step does when the part behind it is unavailable.
If the postcode lookup does not respond, the address fields should open for manual entry instead of waiting. If delivery options cannot be calculated, show the standard option with its price rather than an empty list. If one payment method is unavailable, mark it as temporarily unavailable and keep the others working; a payment step that fails to load because one method did not answer turns a small outage into a lost order. If a payment attempt fails, return the shopper to the payment step with the basket intact and another method in view.
That last state is where reliability and trust meet. A shopper who sees an error after approving a payment in a banking app does not know whether money has left their account. Some try again and pay twice, others leave and contact customer service. A message that says plainly whether the attempt went through, or that the order is being checked and an email will follow, prevents both.
Test these states on purpose. In a test environment, switch off the address lookup, one payment method and the delivery calculation one at a time and walk through an order on a phone. What you see then is what your shoppers will see on your busiest day.
More payment methods turn the layout into a decision
A fallback method only helps if the shopper can find it. Crawford’s contribution to the session was that the final step should feel easy and that no shopper should have to hunt for the method they want, and Maire described offering a frictionless, personalised checkout as payment methods multiply as the challenge ahead.
That multiplication has continued. Since early 2026 Dutch shoppers have seen iDEAL as a combined iDEAL | Wero logo in webshops, so a method they know by sight has changed its appearance on the page. Baymard also reports that 9% of the US shoppers it surveyed abandoned an order because there were not enough payment methods. Adding methods addresses that, but a long row of similar logos can hide the one the shopper came for. The answer is order, not volume: put the methods your own shoppers use most at the top, show names as well as logos, group the rest, and keep the list stable so a returning shopper finds their method in the same place.
Measure the first attempt, not only the completed order
The figure that made headlines from the session was speed: SSENSE had cut the average checkout time on its site by 70%, and Retail Dive reported a direct effect on conversion. Maire called checkout speed and performance “the actual customer experience” for SSENSE. Speed is worth chasing, but it only counts for attempts that succeed. A fast checkout that fails on the first attempt still loses the order, however quickly it failed.
Checkout reports usually count completed orders, which hides how many attempts it took. Add the share of payment attempts that succeed the first time, split by payment method and device, and the number of shoppers who return to the payment step after a failure. A first-attempt rate that drops for one method or at one time of day shows you where the checkout breaks before a shopper tells you. Run the outage test above before your next peak, and fix the step that fails worst first.
Sources: retaildive.com, baymard.com and rabobank.nl
