Every store owner knows the number. Baymard Institute's meta-analysis of 50 separate studies puts average cart abandonment at 70.22% — a figure that has moved less than a percentage point in five years.
The standard response is to buy the customer back: abandoned-cart emails, retargeting, an exit-intent discount. An entire category of software exists to do it.
Here is the thing that category does not want you to notice. Look at what people actually say when asked why they left, excluding the ones who were only browsing:
| Reason | Share |
|---|---|
| Extra costs too high (shipping, tax, fees) | 40% |
| Delivery was too slow | 20% |
| Did not trust the site with card details | 19% |
| Required account creation | 18% |
| Checkout too long or complicated | 17% |
| Website errors or crashes | 17% |
| Unsatisfactory returns policy | 13% |
| Could not see total cost upfront | 12% |
| Card declined | 10% |
| Not enough payment methods | 9% |
(Respondents could select more than one, so these do not sum to 100.)
Now sort them. Two of those lines — extra costs revealed too late (40%) and being unable to see the total upfront (12%) — describe the same failure: the price the customer agreed to was not the price you charged.
That is the largest cause by a wide margin, and it is not a user experience problem. It is a pricing disclosure decision, made by someone in commercial, usually deliberately, usually to make the product look cheaper in a comparison.
So the honest description of the average checkout is this: you are running a two-stage price quote where stage two is a surprise, and then paying for advertising to bring back the people your own quote disqualified.
The abandoned-cart email is a tax
To be clear, recovery flows work and we build them. They are in the scope of every e-commerce project we ship.
But understand what they are. A recovery email is the cost of re-acquiring a customer you already had, who already wanted the product, who had already done all the work. You are paying twice for the same sale — once in acquisition, once in recovery — plus the discount you will probably attach to the second attempt, plus the margin damage of training your customers that leaving the cart produces a voucher.
Order the work accordingly:
- Stop disqualifying buyers (disclosure)
- Stop breaking on them (interface and reliability)
- Stop failing at the last step (payments and operations)
- Then recover the remainder
Most stores do these in reverse, because step four is the one you can buy.
Fixing the 40%
The fix is uncomfortable rather than difficult: show the full, final, delivered price as early as it is knowable.
If shipping depends on destination, ask for the destination at the top of the cart, not at step three. If there are fees, name them on the product page. If tax varies, show the tax-inclusive price for the visitor's region. If you genuinely cannot know until later, say so explicitly with a range.
The objection is always the same: our listed price will look higher than the competition's. Correct. And you will lose some comparison-shoppers at the top of the funnel — cheaply, in a place where losing them costs nothing.
What you get back is every customer who currently reaches your payment screen, discovers a $14 shipping charge, and leaves feeling misled. Those are the expensive losses: fully qualified, fully warmed, fully paid for, and now carrying a negative impression of your brand into the next comparison.
The arithmetic usually favours transparency. Take 10,000 monthly sessions at 2% conversion and a $90 order — $18,000 a month. Assume half your abandonment is cost-surprise driven and you recover a fifth of those people by disclosing early, while losing 5% of your top-of-funnel to the higher visible price:
- Recovered: roughly +7% relative conversion → about +$1,260 a month
- Lost at the top: −5% sessions → about −$900 a month
Marginal in that example, and that is the honest answer at a $90 order value with modest fees. Now run it with a $22 shipping charge on a $60 product, where the surprise is a third of the price, and the trade stops being close.
The point is not that transparency always wins. It is that almost nobody has done the calculation, and the default — hide it and buy retargeting — was chosen by habit.
The 19% who do not trust you
Nineteen percent left because they did not trust the site with their card details. This is the cheapest cluster to fix, because it is almost entirely presentational.
Trust on a checkout page comes from a small number of specific signals: a visible, real company name and address; a working phone number; the payment provider's own branding on the form; a returns policy that is linked rather than buried; no layout shift; no mixed content warnings; a domain that matches the brand; and a page that does not feel slower or uglier than the rest of the site.
That last one matters more than people expect. A checkout hosted on a different subdomain with different typography reads, to a cautious buyer, exactly like a phishing page. If your store's design ends at the cart, you have built a trust cliff at the precise moment you need trust most.
Payments are an operations problem, not a plugin
Two lines — card declined (10%) and not enough payment methods (9%) — are where cross-border stores quietly bleed.
No single provider covers every currency and region. We learned this in detail building Knigopoisk, a store selling Russian-language e-books and audiobooks to readers in many countries. The payment layer is a provider abstraction rather than an integration: Stripe and Polar for cards, YooKassa, Robokassa and T-Bank for rubles, NOWPayments and Cryptomus for crypto. Buyers switch between USD, EUR and RUB, prices recalculate, and only the providers that support the chosen currency are offered. If the default provider is disabled, the next enabled one takes over automatically.
And then the part most stores never build: when a buyer's region has no available payment method at all, the attempt is saved as a lead, with everything staff need to complete the sale manually.
That is the mindset shift. A failed payment is not an error to log. It is a customer who tried to give you money and could not — the most qualified lead your business will generate that day, and in most systems it vanishes without a trace.
The interface 17s
Checkout too long (17%) and site errors (17%) are ordinary engineering.
On length, the rule is that every field must justify itself against the revenue it costs. Company name, "how did you hear about us", a marketing checkbox, a second address line — each one is a small tax, paid by every buyer, forever. Guest checkout removes the single largest structural barrier, addressing the 18% who abandon because you required an account.
On errors, the two that matter are validation and responsiveness. Errors must appear next to the field, name the actual problem, and never clear what the user typed. And the checkout must respond instantly to taps — the interaction metric, not the load metric, is what governs the feel of a form. We go through that in speed is a pricing decision.
The disclosure audit — forty minutes, no tools
Do this yourself, on your own store, on your own phone, with the analytics closed.
1. Pick your most-bought product and add it to the cart. Write down the price you saw on the product page.
2. Note every screen where a new number appears. Shipping, tax, handling, a payment surcharge, a minimum-order fee. For each one, write down at which step it first became visible.
3. Compare the first number to the last. The gap is your surprise. Expressed as a percentage of the product price, that is roughly the size of the problem — on a $60 item with $14 of late-revealed shipping, you are running a 23% price increase at the moment of commitment.
4. Count the fields. Every input between the cart and the confirmation. Then go through them and mark each one as required to fulfil the order, required by law, or we would like to know. The third group is your tax on every buyer, forever.
5. Try to buy without an account. If you cannot, you are choosing to lose a share of the 18% who abandon for that reason in exchange for an email address.
6. Break it on purpose. Enter an invalid card, a wrong postcode, a name with an apostrophe. Does the error appear next to the field? Does it say what is actually wrong? Did it keep what you typed? Most checkouts fail at least one of those three, and each failure lands on a customer who had already decided to pay you.
7. Do it all again on a phone, on mobile data, away from the office. Mobile cart abandonment runs materially higher than desktop, and this step is usually where you find out why.
Forty minutes, a page of notes, and a prioritised list of fixes that cost less than a month of retargeting. We have never run this on a store and found nothing.
What the upside actually is
Baymard's own conclusion, from a decade of large-scale checkout testing: the average large e-commerce site can gain a 35.26% increase in conversion rate through better checkout design alone — which, against combined US and EU e-commerce sales of $738 billion, they put at roughly $260 billion in recoverable orders.
Thirty-five percent. Not from more traffic, not from a new channel, not from a brand campaign. From fixing the last four screens.
That is the whole argument for treating checkout as its own project rather than as the tail end of a redesign. Our e-commerce builds start at $499 with a 1–5 week delivery, and the catalog, cart and checkout are designed as one continuous flow — because the seams between them are where the money goes.
Tell us what you need. You get a fixed price and a delivery date back — free, and with no obligation to proceed.
Frequently asked questions
Is 70% abandonment actually bad?
It is the average, and a large share of it is people browsing with no intent to buy — which is normal and not a problem. What matters is the abandonment among people who reached the checkout, because those are decisions you influenced. Track that number separately; most analytics setups blur the two together.
Should I offer free shipping instead of showing the cost?
Often yes, priced into the product, because it eliminates the largest cause entirely. It is not free — you are moving the cost, not removing it — but a single visible number beats a small number plus a surprise. Test it against margin before committing.
Does guest checkout hurt my customer database?
Marginally, and you can recover most of it by offering account creation after the order is placed, when the buyer has already succeeded and has a reason to want order tracking. Requiring it beforehand trades 18% of your carts for an email address.
How many payment methods do I actually need?
Enough to cover the currencies and regions you sell into, which is a market question rather than a technical one. The more useful design decision is a provider abstraction with automatic fallback, so adding or replacing one later is configuration rather than a project.
Where should I start if I can only do one thing?
Show the full delivered total before the buyer enters any personal information. It addresses the largest documented cause, it costs no engineering time if your cart already calculates shipping, and you will see the effect within a fortnight.
