Checkout is the only part of a store where every visitor has already agreed to buy. That makes it the cheapest place to earn revenue and the most expensive place to be careless — and it makes optimization here overwhelmingly a matter of removal, not invention.
The discipline is simple to state and hard to follow: for every element in your checkout, either it processes the order, or it goes.
Subtract in this order
Delete fields that don't change fulfilment or payment. Company name, second address line, phone number, "how did you hear about us," date of birth. Each is a decision, a keystroke, and — for the fields whose purpose is unclear — a small worry.
Delete the account requirement. Guest checkout first; offer the account on the confirmation screen, when there's an order to track and a reason to want one.
Delete surprises. Any number appearing for the first time inside checkout will cost you more than it would have cost earlier. Move shipping, tax, and fees upstream to the product page or cart.
Delete exits. Checkout doesn't need your main navigation, your promo bar, your chat widget shouting, or your newsletter box. Keep the logo, the steps, and the way back to the cart.
Delete hostility from validation. Validate on blur. Name the fix, not the failure. Never clear a field. Keep the message beside the input.
Then add, sparingly
Three additions consistently earn their place.
Address autocomplete. It removes more typing per line of code than anything else in checkout, and it eliminates a whole class of formatting failures.
Wallet payments. Apple Pay, Google Pay, saved shop credentials. They remove typing and substitute the buyer's trust in the wallet for trust in your form — the two hardest problems at once.
Trust cues beside the card field. Not in the footer. The buyer's question isn't whether your company is real; it's whether this form is safe.
Mobile checkout is a different product
Set proper `autocomplete` and `inputmode` attributes so autofill fires and numeric fields open numeric keyboards. Check that no input hides behind the on-screen keyboard. Make tap targets thumb-sized. Confirm the error message renders where the buyer is looking, not scrolled off-screen.
A checkout that converts on desktop and collapses on mobile has an interface problem, not an audience problem.
Measuring it honestly
Track three transitions, not one: cart → checkout started, checkout started → payment reached, payment reached → order placed. A single "checkout conversion" number averages three separate failures into one uninterpretable figure.
Change one thing. Watch the transition you meant to move. Checkout is the one place where a removed field converts directly into orders, which makes it the best place in a store to practise the habit of changing one thing at a time.
What subtraction can't find
You can see a long form. You cannot see which of its inputs your buyers stop at — and that is the number that decides what to delete next. Two checkouts with the same field count lose people at different fields.
Defrixa's free scan reads your checkout's structure and names its single biggest friction. The field-level answer arrives with the snippet.
Common questions
Neither wins by default. Multi-step works when each step is small and progress is visible; one-page works when the page is genuinely short. Surprises and unnecessary fields hurt both equally.
Offer the ones your buyers use, and at least one wallet. A wall of payment logos becomes its own decision.
Yes, if it tells the truth. An indicator that reveals a fourth step after step three does more damage than none at all.