Chapter 6 of 11
Cart and Checkout: Where Stores Bleed
Cart abandonment is mostly a pricing-surprise problem, not a design problem. What actually causes it, what to fix, and which checkout changes are not yours to make.
Contents
- The abandonment number, honestly
- Surprise costs are most of it
- Cart page or cart drawer
- Every field has a price
- Address entry is the worst part of every checkout
- Taxes, duties, and selling internationally
- Payment and trust at the moment of payment
- Speed and reliability, which nobody tests
- Recovery
- The failure nobody finds: payments that silently decline
- What is not yours to change
By the time someone reaches your cart, they have decided to buy. Everything that happens after that point is you finding ways to lose a customer you already won.
Which makes checkout the highest-leverage surface on most stores, and the most constrained. Some of it you cannot change.
The abandonment number, honestly
Around 65–75% of carts get abandoned. That figure has been roughly stable for more than a decade across every study that measures it.
Do not treat it as a crisis. A large share of those carts were never real purchase intent. People use carts as wish lists, price checkers, and shipping calculators. That behaviour is not a leak, it is how the internet shops.
The number that matters is checkout completion rate: of people who begin checkout, how many finish. Beginning checkout is a much stronger intent signal than adding to cart.
Healthy is usually 45–65%. Below 40% means something specific is broken, and it is almost always one of four things:
- Costs appeared that the customer did not expect
- The form is asking for too much, or asking badly on mobile
- The payment method they wanted was not there
- Something is technically failing on a device or browser you have not tested
Notice that only one of those is a design problem. Checkout optimization is mostly not about layout.
Surprise costs are most of it
Every abandonment study for fifteen years has produced the same top answer: extra costs, shown too late.
Shipping, tax, fees, handling. The customer built a mental total on the product page and the checkout produced a different one. The gap is not usually large. What kills the order is that it was a surprise, because a surprise at the payment step reframes the whole store as untrustworthy.
The fix is not to eliminate shipping cost. It is to eliminate the surprise.
Show shipping cost, or the threshold, on the product page. Covered in Chapter 5, and it belongs there rather than here, because by the time they reach checkout the expectation is already set.
Show it again in the cart, before checkout begins. A cart that shows “Subtotal $64” and nothing else is setting up the failure. Show the estimated total including shipping, even if it is approximate, labelled as an estimate.
If you have a free-shipping threshold, make progress toward it visible. A progress-bar version tends to lift average order value more than conversion rate. That is a legitimate win, and it is exactly the kind of result you only notice if you set a guardrail metric as described in Chapter 3.
Consider whether the threshold is right at all. A $75 threshold on a store with a $48 average order forces a decision on nearly every customer. Sometimes the correct move is not a better progress bar, it is a $60 threshold.
Cart page or cart drawer
A slide-out drawer keeps the customer on the product page and is generally better for stores where people buy several items. A dedicated cart page gives room for order summary, upsells, and reassurance, and is generally better for considered single-item purchases.
Neither is universally right, and this is a poor first test. The gain is small relative to the build cost, and it competes with much larger fixes.
What matters more than which one you use:
- The cart is reachable and its item count is obvious from anywhere
- Quantity edits do not reload the page or lose scroll position
- Removing an item is possible without a confirmation dialog
- Estimated total including shipping appears here, not later
- The primary action is unambiguous and does not compete with three other buttons
That last one is common. A cart with “Continue shopping,” “Update cart,” “Checkout,” and two express payment buttons of equal weight has no primary action. Make checkout visually dominant and demote everything else.
Every field has a price
The strongest generalisation in checkout work: fewer fields, more orders.
Fields most stores can remove or defer:
- Company name. Unless you sell B2B, delete it.
- Address line 2. Make it optional and collapsed behind a link.
- Phone number. Required only if your carrier genuinely needs it. Most do not. If it is for marketing, that is not a checkout field, that is a post-purchase ask.
- Confirm email. Never. Use validation and an inline warning on a suspected typo.
- Account creation. Guest checkout as default, always. Offer the account after the order.
The one we get asked about most is the coupon field. A prominent, empty “Discount code” box tells a customer that a discount exists and they do not have it. Some of them leave to go find one, and a meaningful share do not come back. Collapsing it behind a small text link keeps it available for people who have a code without advertising its absence to everyone else.
On mobile, the mechanics of the fields matter as much as the count. Correct autocomplete attributes so the browser can fill the whole address in one tap. Correct inputmode so the numeric keyboard appears for postal codes and card numbers. Labels that stay visible when the field is filled, not placeholders that vanish. Errors shown inline next to the field, not as a summary at the top that scrolls the customer away from the problem.
Address autocomplete can turn a six-field form into one field, and that change often produces a larger lift than a year of layout work.
Address entry is the worst part of every checkout
The address block is the longest, most error-prone, most abandoned section of any checkout, and it gets almost no attention because it looks solved.
Address autocomplete is the single highest-return checkout change available. One field, type three characters, pick from a list, everything else fills. It removes five inputs, eliminates most typos, and reduces failed deliveries. It costs a small amount per lookup and pays for itself in avoided support tickets alone.
Get the input types right. Postal code fields should bring up a numeric keypad where the format is numeric. Card number fields should be inputmode="numeric" with autocomplete="cc-number". Email fields should not autocapitalise. Each of these is one attribute, and each one is wrong on a large share of stores.
Do not validate too aggressively. Postal code and phone formats vary more than any regular expression you will write. A checkout that refuses a valid international address because it does not match a US pattern loses that order permanently, and you will never see it in your analytics as anything other than an abandonment.
Show errors inline, next to the field, as the customer leaves it. Not on submit, and never as a summary at the top that scrolls them away from the problem. The worst version, still common, clears the entire form on a validation error.
Preserve the entered data through every failure. A payment decline that empties the address fields turns a retry into an abandonment.
Taxes, duties, and selling internationally
If you ship across borders, this section is probably worth more than everything else in this chapter.
Duties collected at the door are the most expensive surprise in ecommerce. The customer pays, receives a demand for another 20% from a courier, refuses the parcel, and you have lost the order, the shipping, and the customer. Delivered-duty-paid checkout, where duties are calculated and charged up front, converts better and generates dramatically fewer support tickets.
Show prices in the customer’s currency, and be honest about conversion. A store that displays in USD to a UK customer and then charges an unexpected amount looks like it is hiding something.
Say where you ship, early. A shipping destination selector that reveals “we do not ship there” at step two of checkout is a customer you wasted.
Local payment methods matter more than most stores assume. iDEAL in the Netherlands, Klarna across much of Europe, Bancontact in Belgium. If a country is a meaningful share of your revenue, its dominant payment method is not optional.
Payment and trust at the moment of payment
Express wallets earn their space. Apple Pay, Google Pay, Shop Pay, PayPal. On mobile they replace a two-minute card entry with a thumbprint. Put them above the card form, not below it, and not hidden behind an accordion.
Beyond three or four options, stop. A wall of payment logos is clutter, and clutter at the payment step reads as an unserious store.
Buy-now-pay-later is a business decision, not a CRO one. It reliably lifts average order value and conversion on higher-ticket items. It also costs 3–6% per transaction. Whether that trade is worth it is a margin question for you, not a test result.
Trust signals at checkout should be functional, not decorative. The padlock in the browser bar does more than any badge you can add. What genuinely helps: your return policy restated in one line near the pay button, a real support contact, and the delivery date repeated. What does not: a grid of security seals from vendors nobody has heard of.
Speed and reliability, which nobody tests
Checkout is the least performance-optimised part of most stores, because it is the part nobody shows off and the part least likely to be measured by whoever runs page-speed reports.
It is also the part where a failure costs the most, because the customer is one tap from paying.
Measure checkout page speed separately. Site-wide averages are dominated by product pages. Pull checkout out and look at it on its own, on mobile, on a throttled connection.
Audit third-party scripts in checkout specifically. Every analytics, chat, personalization, and survey tool somebody has ever added. Each one is a network request between your customer and their order, and a share of them are for tools nobody has looked at in a year. This is the highest-yield cleanup available and it takes an afternoon.
Watch what happens when a third party fails. Not if. When. If your review widget or chat provider has an outage, does checkout still work? Load anything non-essential asynchronously so a slow vendor cannot block a purchase.
Instrument the errors. Most stores have no idea how often a customer sees a checkout error, because errors do not fire conversion events. Log them with enough context to reproduce, and read the log weekly. A single browser-specific failure can quietly cost a few percent of orders for months.
Recovery
Some abandonment is unrecoverable. Some is a customer who got interrupted.
Email recovery is the highest-return automation in ecommerce. Three messages: one at roughly one hour, one at 24 hours, one at 72. The first should be plain and helpful, not a discount. Most recovered revenue comes from the first email, and leading with a discount teaches your customers to abandon carts deliberately.
Do not discount in message one. If you must discount, do it in message three, and make it small. A store that reliably sends 15% off an hour after abandonment has not built a recovery flow, it has published a coupon and taught its customers where to find it.
SMS works and is easy to overdo. One message, well-timed, with genuine consent. It converts better than email and annoys faster.
Exit-intent on the cart page is a last resort. It works on desktop, does not exist meaningfully on mobile, and it is the most intrusive thing you will do to a customer. If you use one, offer help rather than a discount: shipping information, a size guide, a support link.
Measure recovery revenue net, not gross. Recovery reporting is the most flattering number in ecommerce, because most tools credit the full order value to the email whenever a recovered customer eventually buys. A share of those people were coming back anyway. The honest version is a holdout: withhold the sequence from a small random slice of abandoners for a month and compare. The result is always smaller than the dashboard claims, and it is the number to plan with.
The failure nobody finds: payments that silently decline
Somewhere between 5% and 15% of card attempts fail on a typical store. Most operators have never looked at this number, because it does not appear in a conversion funnel. The session simply ends.
Worth checking, in order:
Your decline rate, by card type and by country. Your payment provider reports it. A rate materially above 10% points at either a fraud filter set too aggressively or a mismatch between where you sell and what you accept.
Your own fraud rules. Fraud tooling is usually configured once, defensively, and never revisited. A rule that blocks orders where billing and shipping countries differ will block a large share of legitimate gift purchases. Pull a sample of blocked orders and read them; a meaningful share are usually real customers.
3D Secure friction. Strong customer authentication is mandatory in some markets and optional in others. Where it is optional and applied to every transaction, you are adding a step and a failure mode for no gain.
What happens after a decline. This is the part that is nearly always broken. The customer should see a clear, non-accusatory message and a way to try another method, with all their data still in place. What they often get is a generic error, an emptied form, and no path forward.
A store doing $4M with a 12% decline rate and a broken retry path is losing a six-figure sum a year to a screen almost nobody on the team has seen.
What is not yours to change
If you are on Shopify, the checkout is largely not yours. On Plus you get more control through checkout extensibility, and on standard plans you get considerably less. Fighting this is a waste of a quarter.
Which is genuinely good news, because it forces the effort to where it belongs. The pre-checkout experience is where your leverage is, and it is fully yours:
- Cost transparency on the product page and in the cart
- Delivery expectations set before checkout begins
- Express wallets enabled and ordered correctly
- The cart itself, which you fully control
Do not re-platform for checkout conversion. We have watched stores spend six figures and two quarters moving platforms to gain checkout control, and land roughly where they started because the actual problem was a shipping threshold set three years ago and never revisited.
And do not test your checkout on a store that cannot support it. Checkout traffic is a fraction of site traffic, so the sample size arithmetic from Chapter 4 bites hardest here. Run the numbers in the significance calculator using checkout starts as your baseline population, not sessions. Most stores discover their checkout is untestable and their correct move is to fix the obvious things and ship.
Next: landing pages for paid traffic, where the rules are different because the visitor arrived with a promise in their head. If you want the whole funnel walked and ranked rather than described, that is the engagement.
What founders ask before they book.
What is a normal cart abandonment rate?
Roughly 65-75% of carts are abandoned across most ecommerce categories, and it has been in that band for over a decade. Treat a number in that range as normal rather than as a crisis. What matters more is your checkout completion rate, which is the share of people who begin checkout and finish it. Healthy is usually 45-65%, and a number below 40% points at a specific fixable problem rather than at general internet behaviour.
Should I use a one-page checkout or a multi-step checkout?
The number of pages matters much less than the number of fields and the point at which costs appear. Multi-step checkouts often outperform one-page ones because each step feels short and progress is visible. Do not re-platform your checkout to change page count. Reduce fields and show shipping cost earlier instead.
Does offering more payment methods increase conversion?
Adding the express wallets your customers already use, typically Apple Pay, Google Pay, Shop Pay and PayPal, is one of the more reliably positive changes in checkout, because it removes manual card entry on mobile. Adding a long tail of niche methods beyond that produces little and adds visual clutter. Two or three well-placed express options is the useful version.
Should I require account creation at checkout?
No. Guest checkout should be the default and the visually primary path. Offer account creation after the order is placed, when you already have the email and the customer has a reason to want order tracking. Forcing account creation before payment is one of the most consistently damaging things a store can do to its checkout.