Skip to content
UXE Award

Checkout UX: what separates a good e-commerce flow from a scoring one

The checkout details that decide an e-commerce UX score: guest checkout, address entry, error recovery, honest totals, and the delivery promise. What editors test, in the order they test it.

The UXE editors5 min read

Checkout is the most scrutinised flow in digital design and still the most commonly broken. Editors scoring an E-commerce Experience entry go through it with a card in hand, on a phone, usually as a guest — and most of what separates a good flow from a scoring one happens in details nobody photographs.

The order editors test it in

Four passes through the same checkout
  1. 01

    As a guest

    On a phone. No account, no autofill, cold.

  2. 02

    With a mistake

    Wrong card number, wrong postcode, a required field skipped.

  3. 03

    With the keyboard

    Tab through the whole flow. Then a screen reader.

  4. 04

    On a bad connection

    Throttled. Does a double tap create two orders?

1. Guest checkout, offered first

Forcing account creation before purchase remains the most expensive pattern in e-commerce. The strongest flows offer guest checkout as the primary path and then offer to save the details after the order is placed, when the customer has a reason to say yes and the work is already done.

If your business genuinely needs accounts, say why at the moment you ask — "so you can track delivery" earns more consent than a bare registration wall.

2. Address entry that respects the world

Address forms are where international customers discover whether they were thought about. Anyone testing from outside your home market finds this in about forty seconds.

Address forms, two ways
Compared onScores wellCosts points
FormatAdapts to the selected countryUS-shaped form for every country
PostcodeOptional where it doesn't existRequired, numeric, 5 digits
LookupOffered, with manual entry always availableLookup only, no way to type an address it doesn't know
NameOne full-name field, or clearly labelled partsFirst/last assumptions that break half the world's names
PhoneInternational format accepted, explained why it's neededRejected unless it matches one country's pattern

3. Errors that can be recovered from

Everyone tests the happy path. Editors test failure, because failure is where design shows. The benchmark is simple: after an error, can the customer fix it without losing anything or re-reading the whole page?

  • Errors appear next to the field, not only in a summary at the top — and the summary, if present, links to the field.
  • Nothing already entered is cleared. Especially not the card field, and especially not on mobile.
  • The message says what to do: "Your card was declined by your bank — try another card or contact them" rather than "Transaction failed (code 51)".
  • Validation fires when the field is left, not on every keystroke, and not only after submit.
  • A failed payment returns to a checkout that still has the basket, the address and the delivery choice.

4. The total is honest, early

Surprise costs at the final step are the most cited reason for abandonment, and they read as a decision made against the customer. Delivery, tax and fees belong on the basket, with a real figure or an honest estimate and the reason for the estimate.

Any total that changes after the customer has decided to buy is a design that treats its own pricing as a thing to be hidden. Editors score that under ethics, not just usability.
From the Inclusive & Ethical Design brief

5. The delivery promise is specific

"Ships in 3–5 business days" is a shipping policy. "Arrives Thursday 26 September if you order in the next four hours" is a design decision that removes a question from the customer's head. Products that compute a real date, and stand by it, score higher on usability and on trust — and the difference costs one API call.

6. It doesn't double-charge on a bad connection

Tap pay, nothing happens, tap again. Every editor tries this. The flow must disable the control on first submit, show that something is happening within 100ms, and be idempotent on the server. A checkout that can create two orders is a correctness problem wearing a UX costume.

How often each appears in entries that clear 8.0
  • Guest checkout95%Offered as the primary path in nearly every high-scoring entry.
  • Field-level errors90%
  • Total shown on basket85%
  • Real delivery date60%The clearest differentiator between good and excellent.
  • Idempotent submit55%Rarely mentioned in entries; always tested by editors.
An impression from reviewing, not a measured study — read it as what editors keep noticing, not as a statistic.

What to send if you're entering a checkout

  • A working test path, or a sandbox card that lets us complete a real purchase.
  • The error state after a declined card, on mobile.
  • The basket showing the full total, including delivery.
  • The order confirmation — including what it says about returns, which most teams forget is part of the flow.
  • One measured outcome: abandonment, completion time, or support contacts about orders.

Checkout entries usually belong in E-commerce Experience, often alongside UX Writing & Content Design — because most of what makes a checkout calm is its words. Enter yours.

Payment methods and the wallet question

Express wallets shorten checkout dramatically, and most stores implement them in a way that undoes the benefit: offered after the address form rather than before it, or offered so prominently that customers who want to pay by card can't find the option.

  • Offer wallets first, before any form, where they can skip the form entirely.
  • Keep the card path visible and unapologetic — it is still the majority in most markets.
  • Show accepted methods on the basket, not only at the final step.
  • For a market with local methods (bank transfer, direct debit, local schemes), surface them by region rather than burying them under "other".

The mobile keyboard details that decide a conversion

An editor filling in your checkout on a phone notices these within seconds, and almost every store gets at least one wrong.

Five fields, on a phone
Compared onRightWrong
Card numberNumeric keypad, autofill enabled, spacing as you typeFull keyboard, autofill blocked
ExpiryNumeric, auto-advance, accepts 04/28 and 0428Two dropdowns
PostcodeAppropriate keyboard for the countryNumeric everywhere
EmailEmail keyboard, no auto-capitaliseAuto-capitalised first letter
AutofillCorrect autocomplete attributes throughoutNone, or wrong ones

Questions we get asked

  • Should checkout be one page or several steps?

    Either works. What matters is that progress is visible, nothing is lost when moving backwards, and each step has a single clear purpose. Multi-step checkouts fail when they hide the total; single-page ones fail when they overwhelm on a phone.

  • Is guest checkout still necessary?

    Yes. Forcing account creation before purchase remains the most expensive pattern in e-commerce. Offer to save the details after the order instead, when the work is already done.

  • How do editors test a live store?

    With a real or sandbox card, as a guest, on a phone, and then again with a deliberate mistake. Give us a test path in the entry if you'd rather we didn't place a real order.

  • Do countdown timers and stock warnings count against us?

    Only when they aren't true. A real deadline stated plainly is fine. A timer that resets on reload, or "3 left" that is permanently 3, is scored as a dark pattern under ethics.

Assessed work in commerce

Chosen by score from the work the editors have assessed, not by hand.

  • ecommerce
  • checkout
  • conversion

Written by

The UXE editors

The team that selects the work, uses it, scores the seven criteria and writes every verdict on this site.