Checkout advice tends to arrive as a list of layout changes: fewer steps, a progress bar, trust badges, a bigger button. Some of it helps a little. None of it addresses where most of the money actually goes.

The large losses are informational and operational rather than visual. They are things the customer discovers at the worst possible moment, and they are fixable without redesigning anything.

1. The total appears too late

Unexpected cost at the final step is consistently the largest single cause of abandonment in published research, and it is the one most stores create for themselves. Shipping calculated on the last screen, a handling fee, tax added at the end, a card surcharge.

The fix is not to remove the cost. It is to stop it being a surprise: show the delivery cost, or a usable estimate, on the product page and in the basket. A customer who sees the real total early either accepts it or leaves early, and leaving early costs you nothing but a session. Leaving at the payment step costs you the whole funnel.

2. Forced account creation

Requiring registration before purchase is asking someone to commit to a relationship before they have bought anything. It exists because the business wants the data, not because the customer wants an account.

Guest checkout, with an optional "save these details" after the order is placed, gets you most of the accounts and none of the abandonment. The customer has already typed everything; creating the account is one checkbox at a point where they are pleased with you rather than impatient.

3. The delivery answer is missing

Not the price - the answer. When will it arrive, can I choose, what happens if I am out, can I collect, do you deliver to my country at all.

This is the question that sends people to a competitor to find out, and they frequently buy there rather than coming back. A date, or even a range, is worth more than a discount: it converts "probably" into a decision.

For B2B the equivalent question is lead time on non-stock items, and the answer "we will contact you" is functionally the same as no answer.

4. Payment methods your customers actually use

This is market-specific and easy to get wrong from the inside. A store selling into a market where bank transfer or a local wallet is normal, offering only cards, is not losing conversion rate - it is losing entire customer segments invisibly.

The check is boring and effective: look at where your traffic comes from, find out what the normal payment method is there, and see whether you accept it. Then check that the wallet options appear as buttons early, not as a radio option buried under card fields.

5. Form and validation failures

This is the part that is genuinely a build problem, and it is where checkouts quietly leak on mobile.

  • Validation that rejects a valid phone number, postcode or address format because it was written for one country.
  • Errors reported only after submit, at the top of the page, with the offending field somewhere below the fold.
  • A form that clears itself on an error - the fastest way to lose someone who has already typed an address on a phone.
  • Missing autocomplete attributes, so the browser cannot fill anything and every field is typed by hand.
  • Wrong input types, so a phone number gets a full alphabetic keyboard.
  • A layout shift as a field appears or an error renders, moving the button under a thumb that has already committed.

None of these are hypotheses that need testing. They are defects. Fix them and measure afterwards.

6. Nothing answers the doubt

At the payment step people are deciding whether to trust you with money. What answers that is specific and dull: a returns policy in plain language, a real address, a phone number that is answered, a delivery guarantee with a consequence attached.

What does not answer it is a badge. Generic trust seals are visual noise; several are recognisable to nobody outside the industry that sells them.

Where to start

Instrument the steps first. Most analytics setups can tell you the checkout completed and nothing about where the other people went. Without step-level events, every conversation about checkout is opinion.

Then segment by device before you change anything. Desktop and mobile checkouts usually fail for different reasons, and an aggregate rate hides both.

Then fix the defects in section five, because they need no hypothesis. Only after that is it worth arguing about layout - and by then you will have the data to argue with.