For nine months I watched people pay for things. Not metaphorically: forty recorded sessions of Northwind customers in the last minute before the order, played back in a meeting room with the sound on.

What I expected

I expected the payment page to be the problem. It looked old, it looked like a different company, and every designer who saw it wanted to fix it.

What I saw

People slowed down earlier. They slowed down on the address form, which had seven fields and no lookup, and they slowed down when the delivery fee appeared for the first time. The payment page, old as it was, did not lose many people. By the time they reached it they had decided.

The lesson

The last step of a funnel is where every earlier promise is kept or broken. A product page that promises free delivery and a checkout that adds a fee is not a checkout problem; it is a trust problem that shows up at the checkout.

• Show the fee on the product page
• Replace the seven fields with a lookup
• Then, and only then, redesign the payment page

The number

Conversion went from 2.9 % to 3.8 %. The address lookup alone was worth more than the payment redesign that everyone wanted first.

Trust is spent at the last step and earned at every step before it.

If you are about to redesign your checkout, watch forty sessions first. The problem is probably earlier than you think.