Checkout is a product, not a plugin
Most conversion work happens on product pages. Most conversion is lost after the customer has already decided to buy — in the last four screens.
By the time a customer reaches checkout they have decided to give you money. Everything that happens next can only lose the sale. That asymmetry should make checkout the most instrumented, most carefully designed surface in the whole store, and in most businesses it is the least.
The largest single lever in Europe is payment method coverage, and it is not close. A Dutch customer expects iDEAL, a Belgian one Bancontact, a German one an invoice option or Klarna. Offering cards only in those markets is not a design decision, it is a decision to lose a share of every basket.
The second lever is not asking for things you do not need. Every optional field costs completions. Company name, second address line, phone number — each one needs to justify itself against a measurable drop, and most cannot.
Guest checkout is the third, and it is still an argument we have with clients who want the account. You can ask for the account after payment, when the customer has a reason to say yes. Asking before it is a toll gate.
None of this can be argued from opinion, which is why we instrument every step before changing anything. Knowing that eleven percent of customers abandon on the shipping step is a fact. Believing the button should be green is not.
On one forty-market retailer these changes moved checkout conversion by thirty-one percent in a quarter. The product pages were untouched.