Most optimisation money goes to the top of the funnel. The expensive leaks are in the last five steps, where the customer already decided to buy and something stopped them.
Almost every euro spent on improving a store goes to the top of the funnel: more traffic, better ads, another campaign. Meanwhile the most expensive losses happen in the last five steps, among people who had already decided to give you money and then stopped.
Those customers are the cheapest ones you will ever have. You already paid to acquire them. Recovering a fraction of them costs nothing in media budget, which is why this is the first place we look when a store is underperforming.
There is no shortage of advice about what a good checkout looks like. The problem with a best-practice list is that it has no priority in it — and priority is the only thing that matters when you have a fixed budget and a long list.
Use a crude number instead: how many people reach a step, what share drops there, and what an order is worth to you in margin. A step where a fifth of buyers vanish at two hundred orders a month is worth ten times more attention than a step everybody on the internet says is important. Do the arithmetic before you do the work; it usually reorders the whole list.
People rarely abandon because a page is ugly. They abandon because they still have a question and cannot be bothered to ask it: when does it arrive, what is the total with shipping, will it fit, what happens if it is wrong, is this the right size for my situation.
You do not need research to find your own version of this. Read your support inbox for one week and count the repeated questions. Every question a customer had to ask is a line missing from the page — and the customers who asked are the minority. The rest simply left.
Shipping and tax revealed at the third step of checkout is the single most reliable way to lose someone who was ready to pay. It is not the amount that does the damage, it is the surprise: the price they had accepted turns out to be a different price, and the whole transaction feels slightly dishonest.
Show the real total as early as you can, even when it looks worse on the product page. A store that quotes twelve euros of shipping on the product page converts better than one that quotes it after the address form, because the first one only loses the people who were never going to buy at that price anyway.
Forced account creation is still one of the largest self-inflicted losses in e-commerce. So is the phone number you never call, the company field for consumers, the second address line, the newsletter checkbox that has to be understood before the payment button.
Go through your own checkout with an unfriendly question at every field: which department uses this, and what happens if it is empty? Fields that survive that question stay. The rest are a tax you charge yourself on every order.
This one is invisible in your analytics, which is exactly why it survives for years. There is no event for “I did not see the way I pay for things”. The customer simply leaves, and the drop looks like ordinary indecision.
Check what people in your markets actually use — the answer differs sharply by country, and the local method is often the majority rather than a nice extra. If you sell across borders, this is frequently the largest single number in this article. We wrote about the payment step in detail in a separate piece, because it deserves its own list.
Store speed is usually discussed as an average, and averages hide the thing that hurts. What matters is the specific step where the spinner lives: the payment redirect, the address validation, the shipping calculation that queries a carrier, the page that recalculates the basket after a discount code.
Those moments arrive when the customer has already committed, which makes them the worst possible place for a delay — the person is holding a card and looking for a reason to abandon a decision they are slightly unsure about. Our latency budget for checkout covers how to put numbers on this, and the third-party tax covers who is usually responsible for the delay.
There is a hesitation that happens with the card already in hand, and it has nothing to do with usability. The buyer does not know you. They are about to send money to a company they found forty seconds ago, and they are looking for a reason to believe the parcel will arrive.
What closes that gap is unglamorous and cheap: a real company name and address somewhere findable, a returns policy written in plain language rather than legal defence, a delivery promise with a date in it, and a way to reach a human. Reviews help, but they help less than the absence of obvious reasons for doubt.
This leak is largest for stores with paid traffic and small brands, because those visitors arrive with no prior relationship at all. If you sell across borders, add one more: a page that makes it obvious you ship to their country and what that costs, before they have to guess.
The thank-you page and the confirmation email are part of the purchase, not the aftermath. A confirmation that arrives an hour late, lands in spam or contains none of the information a nervous buyer wants produces support tickets, chargebacks and refund requests — a cost that shows up in a different budget line and therefore never gets attributed to checkout.
Send it immediately, include what happens next and when, and make it survive being read on a phone. This is the cheapest customer-retention work available to any store.
Take the two or three worst numbers and put them next to each other with the arithmetic visible: this many people reach the step, this share leaves, an order is worth this much in margin. Now the list has an order, and it will rarely match the order you would have guessed. A payment method that affects a quarter of one market can outrank a redesign of the entire product page, and it takes an afternoon.
Then pick a single number to watch weekly, and make it checkout completion split by device rather than overall conversion. Overall conversion moves with your traffic mix and tells you almost nothing about whether the work helped; completion rate by device responds to exactly the changes described here.
Rank what you found by expected recovery against effort, and be honest that the top of the list is usually dull: a sentence added to a product page, a cost shown earlier, three form fields deleted, one payment method enabled. None of it will look like a redesign. All of it will be measurable within a month, which is more than most redesigns can claim.
The worst step in your funnel is almost always the one you have never seen, because you have never bought from your own store the way your customers do. Start there and the rest of the list gets shorter on its own.