The Checkout That Loses 35% of the Revenue: A Breakdown of the Baymard Data

Every ecommerce conversation eventually arrives at the same number. Roughly seven carts in ten are abandoned. In a pitch deck it becomes evidence that something is badly broken. In a board meeting it becomes evidence that nothing can be done, because everyone loses seven in ten. Both readings are wrong, and the gap between them is worth a great deal of money.

The number itself is sound. Baymard Institute maintains a running average of documented cart abandonment across the published research, and as of its last update on 22 September 2025 it sits at 70.22%. What matters is not the headline but what sits underneath it: a ranked list of abandonment reasons that are mostly within a store owner’s control, a benchmark showing that almost no checkout on the open internet is actually good, and an estimate of how much conversion an average large store is leaving on the floor.

This article takes that data apart. Which parts are solid, which parts are routinely misquoted, and what any of it should change on Monday morning.

What 70.22% is, and what it is not

The figure comes from Baymard’s running list of cart abandonment statistics, and it is explicitly an average calculated across 50 different published studies. That construction has three consequences that get lost in the retelling.

• It is a meta-average, not a measurement. Fifty studies means fifty definitions of what counts as a cart, fifty sampling windows, fifty industry mixes. The average is a useful orientation point and a terrible target.
• It includes people who were never going to buy. Adding to cart is how a large share of shoppers save items, compare prices and check delivery cost. That behaviour is not a failure of your checkout, and no amount of interface work removes it.
• Your own rate is the only one that matters. A store selling GBP 400 furniture and a store selling GBP 12 consumables should not have the same abandonment rate, and if they do, one of them has a problem the benchmark cannot see.

The right use of 70.22% is as a sanity check. If your measured rate is far above it, you have a specific defect worth hunting. If it is far below it, check your tracking before you celebrate, because an abnormally good number is usually an instrumentation artefact.

The reasons, ranked

The more useful half of the same dataset is the survey of why people abandoned during checkout, deliberately excluding respondents who said they were only browsing. That exclusion is what makes the list actionable: these are people who intended to buy and then stopped.

The percentages add up to well over 100 because respondents could select more than one reason. Read them as prevalence, not as a share of lost revenue.

Two things stand out. First, the single largest cause is not a design problem at all: it is a pricing presentation problem. Forty percent of intent-to-buy abandonment is people discovering a cost late. Second, the four interface-level reasons together (account creation, length, errors, no visible total) describe a checkout that most teams could measurably improve inside a quarter without touching the commercial model.

Where the 35% actually comes from

The number that gets used to justify checkout budgets deserves a careful reading. Baymard’s cart and checkout usability research benchmarked 344 top-grossing US and EU ecommerce sites against 110 or more guidelines, with over 30,000 checkout elements manually reviewed and scored. The findings are blunt:

• 65% of sites perform mediocre or worse.
• Only 2% reach a rating of good.
• None reached a state-of-the-art experience. Not one.
• The average site has 32 distinct improvements available in its checkout flow.

The 35% is the conversion uplift Baymard’s combined usability test sessions suggest an average large-scale ecommerce site could potentially reach by fixing those 32 things. That phrasing matters. It is a modelled ceiling derived from qualitative testing, not the measured average outcome of real redesigns. Nobody should sign a contract that promises it, and any agency quoting it as a guaranteed result is quoting it dishonestly.

A later Baymard round, summarised in their overview of the current state of checkout usability, put 64% of 180 or more leading ecommerce sites at mediocre or worse on checkout UX. The two samples differ, but the conclusion has been stable for years: the median checkout on the internet is not good, which is precisely why the upside is large.

Check the vintage of your form-field benchmark

Here is a figure worth retiring. The claim that the average checkout has 14.9 form fields circulates widely and is out of date. Baymard’s own tracking of average checkout field counts puts the average at 11.3 fields in 2024, down from 11.8 in 2021 and 12.7 in 2019. The 14.9 measurement dates from 2016. Quoting it in 2026 overstates the problem by roughly a third and makes the rest of your analysis easy to dismiss.

The recommendation has not moved: most stores can complete an order with about 8 fields. So the real gap is roughly three fields, not seven. That is still worth closing, but it changes the shape of the work from a rebuild to a triage.

One further caution. A widely repeated line says every field beyond the eighth costs 4 to 6 percent of completions. We could not trace that figure to Baymard’s own published field-count research, and we are not going to present it as though we had. Treat it as folklore until someone shows the study.

The three fixes that map to the three biggest reasons

Move the true total forward. If 40% of intent-driven abandonment is cost shock, the fix is not a discount, it is disclosure timing. Delivery cost estimated on the product page, a running total visible throughout, no fee that first appears at the payment step. The temptation is to solve this by hiding the cost even later, which is the exact pattern we argued against in our piece on where persuasion turns into manipulation. It converts once and it churns forever.

Let people buy without an account. Eighteen percent is an enormous amount of revenue to trade for a database row. Offer account creation after the order is placed, pre-filled from what they just typed. Conversion to registered customer is typically higher there anyway, because the buyer has already received something.

Make the form disappear. Correct autofill tokens in the HTML specification are free to implement and cut typing dramatically on mobile. The payment and address form guidance on web.dev is the practical companion. The same logic that makes a lead form convert applies here, and we went through it in detail in our guide to multi-step forms and adaptive fields.

What the data does not justify

• Budgeting for the 35%. It is a potential ceiling on an average large-scale site. Plan for a measurable improvement, and treat anything above ten percent as a good year.
• Blaming the checkout first. If traffic arrives unqualified, the checkout is the last place to look. Our breakdown of why a site produces no enquiries covers the upstream causes that get misattributed to the final step.
• One-shot redesigns. Baymard’s own framing is 32 discrete improvements, not one project. Sequential fixes with measurement between them beat a big-bang relaunch, which changes fifty variables at once and teaches you nothing.
• Copying a competitor’s checkout. Given that only 2% of benchmarked sites rate as good, the odds that your competitor is one of them are poor.

How to get your own number

Before any of the above, instrument the funnel properly. Four events, measured separately on mobile and desktop: cart viewed, checkout started, payment attempted, order completed. The drop between any two of them localises the defect faster than any benchmark. If the fall is between cart and checkout start, it is a cost or trust problem. Between checkout start and payment attempt, it is a form problem. Between payment attempt and completion, it is a payments problem, and no amount of design work will touch it.

Read the result against the direction of travel. Across 2026 the pattern on measured panels has been fewer sessions and more revenue per session, which we set out in what actually happened to ecommerce in 2026. If your traffic is down and your checkout completion is flat, you are living that pattern and the conversion work is where the return is. If completion is falling as well, something in the flow broke and the market is not your problem.

Then take the number of orders you complete in a month, apply the abandonment rate you actually measure, and price a single percentage point of recovery. On most stores that arithmetic ends the debate about whether the work is worth doing.

The short version

The 70.22% average is a meta-figure across 50 studies, updated in September 2025, and it is a sanity check rather than a target. The abandonment reasons are the actionable part: cost shock at 40%, forced accounts at 18%, complexity at 17%. The 35% conversion figure is a modelled ceiling from usability testing across 344 benchmarked sites where only 2% rate as good and none rate as excellent. And the average checkout carries 11.3 fields against a need for roughly 8, not the 14.9 you will still see quoted.

A checkout that leaks at every one of those points is usually not a checkout problem but a platform decision made years earlier. That is the sort of thing we untangle when we build and rebuild online stores, and it is visible in how we approached the catalogue and order flow for Fense and the purchase path for Lunares.