The number that has not moved in a decade
The Baymard Institute maintains a running meta-analysis of cart abandonment, drawn from around 50 published studies and weighted by study quality. The current aggregate is 70.22%.
Individual studies range from about 55% to 84%. The average has sat near 70% for more than ten years, through every wave of platform improvement, one-click payment, and design convention.
A useful companion figure: Dynamic Yield measures around 77.8% across more than 200 million monthly users on a rolling basis. The two do not conflict. Baymard's is a long-run aggregate across many studies; the live measurement runs higher because it captures real sessions in real time. Use the first as a durable benchmark and the second as a behavioural reading.
One qualification that most articles omit: the floor is not zero. Comparison shopping, tab-based research, and pure browsing mean a share of cart additions were never going to convert regardless of how good the checkout is. The realistic target is a better number, not a solved one.
Why this is the one part of Drive where effects are large
Behavioural nudges produce around 1.4 percentage points at scale. Checkout work produces far more, and the reason is worth understanding rather than treating as a lucky exception.
Nudges are usually persuasion, applied to people who have not decided. Checkout optimisation is friction removal, applied to people who have already decided and are actively trying to give you money.
In COM-B terms, most abandonment is an opportunity problem rather than a motivation problem. The customer wants to buy and something is preventing it. Removing an obstacle from a motivated person is the highest-yield intervention available in marketing.
The 35% claim, graded
Baymard's headline finding is that the average large e-commerce site can increase conversion by 35.26% through checkout design improvements alone, which they extrapolate to around $260 billion in recoverable orders across US and EU e-commerce.
Under the grading used across this series, this is a documented practitioner method with a vendor interest. Baymard sells checkout usability research and consulting, and the figure is their own. It is also better documented than most vendor claims — derived from years of large-scale usability testing on live production sites with real users, not from a customer survey.
Treat 35% as an upper bound available to sites with substantial unfixed problems, not as an expected return. The direction is well supported. The specific number describes what the worst-performing large sites could recover.
What actually causes abandonment
The research is unusually specific, which makes it actionable.
Unexpected costs, cited by around 48% of abandoners. Shipping, taxes, and fees revealed at checkout rather than earlier. This is the single largest cause and among the most fixable — the fix is disclosure timing, not discounting.
Too long or complicated a process, cited by nearly one in five. Baymard's benchmarking finds the average checkout uses around 23 form fields where 12 to 14 is achievable, and that most checkouts could cut displayed form elements by 20 to 60%.
The volume of unfixed issues. Their benchmark of leading e-commerce sites found an average of 39 potential checkout improvements per site — including at large companies that had already run optimisation programmes.
Mobile. Abandonment on mobile runs roughly 73 to 80%, close to ten points above desktop. Mobile now carries the majority of e-commerce traffic while converting at well under half the desktop rate. For most businesses this single gap is the largest available conversion opportunity.
Use the funnel as accounting, not as psychology
The funnel is a poor description of how people buy. Purchasing is non-linear, involves comparison across sessions and devices, and mostly happens among people who were not in market when they first encountered the brand.
It is an excellent structure for locating leaks.
That distinction determines how to use it. Do not design messaging around funnel stages as though customers move through them in order. Do instrument every step so that drop-off is measurable and attributable to a specific point.
The practical method is unglamorous: measure the conversion rate between each consecutive step, find the largest absolute loss, and work there. Not the step that feels most improvable, and not the page the team most wants to redesign.
Absolute matters more than relative. A step converting at 20% with a million visitors is a larger opportunity than a step converting at 2% with a thousand.
Where conversion work goes wrong
Tests are underpowered. The most common and most damaging error. Most conversion tests lack the sample to detect the effects they are looking for, which means results reflect variance. This is the same problem that affects pilot design generally, and it applies with full force here.
Local maxima. Iterating on button colour and headline wording optimises within an existing structure. When the structure is the problem — a required account, a missing payment method, an unclear delivery promise — no amount of refinement reaches it.
Optimising traffic that should not have been bought. A funnel converting badly because the audience was wrong is an acquisition problem being addressed at the wrong end.
The cause sits below the visible step. Customers abandon where they encounter the symptom, not where it originates. A refund policy nobody can find, a stock accuracy problem, a payment method that fails silently — the checkout page reports the loss and did not create it.
Discounting instead of disclosing. The response to cost-driven abandonment is usually a discount. The evidence points at surprise rather than price: costs revealed late. Showing them earlier costs nothing and addresses the actual mechanism.
Diagnostic: is conversion work targeting the right thing?
Six tests.
Step-to-step conversion is measured, and the largest absolute loss is known.
Delivery costs are visible before the checkout, not revealed at the payment step.
The number of form fields has been counted and compared against what is genuinely required.
Mobile and desktop are measured separately, with mobile treated as the primary case.
Tests are sized to detect the effect being sought, and duration is fixed in advance.
At least one abandonment cause has been traced to an operational process rather than a page.
Test six is the one most teams have never attempted, and it is where the durable improvements are.
What this produces
Revenue from demand you have already paid for.
That is the specific value, and it is why conversion work is usually the first thing to fix rather than the last. Every improvement here raises the return on all upstream spend simultaneously, which no channel optimisation can do.
The proportion is worth holding onto. Seven in ten people who reach a cart do not buy. Some of those were never going to. A meaningful share were, and stopped because of a cost revealed too late, a form too long, or a step that failed on a phone. Those are cheap problems producing expensive outcomes, and they are the most reliably solvable thing in this stage.
Frequently asked questions
What is the average cart abandonment rate?
Around 70.22%, from Baymard Institute's meta-analysis of roughly 50 studies. Live-session measurement runs higher — Dynamic Yield reports around 77.8% across more than 200 million monthly users. The figure has been stable near 70% for over a decade.
Can cart abandonment be eliminated?
No. Comparison shopping, tab-based research, and browsing intent mean a portion of cart additions will never convert. The realistic goal is reducing the fixable share.
How much can checkout improvements deliver?
Baymard's research indicates up to 35.26% conversion improvement for the average large site. Baymard sells checkout research, so treat that as an upper bound for sites with substantial unfixed problems rather than an expected return.
What is the biggest cause of abandonment?
Unexpected costs, cited by around 48% of abandoners. The mechanism is surprise rather than price, so the fix is showing shipping, taxes, and fees earlier — not discounting.
How many form fields should a checkout have?
Baymard's benchmarking finds the average checkout uses around 23 where 12 to 14 is achievable, with most checkouts able to reduce displayed elements by 20 to 60%.
Is mobile really that different?
Yes. Mobile abandonment runs roughly 73 to 80%, close to ten points above desktop, while mobile carries most e-commerce traffic. For most businesses it is the single largest conversion gap.
Why do our conversion tests never reach significance?
Usually because the sample is too small for the effect size sought. Sizing the test from a minimum detectable effect, set on commercial grounds, is the fix — and some effects are too small to detect at your traffic level.
Sources
Baymard Institute, cart abandonment meta-analysis across approximately 50 studies, and checkout usability benchmarking — a provider of checkout usability research and consulting
Dynamic Yield, rolling live-session abandonment measurement across 200 million-plus monthly users
Published device-level and regional abandonment benchmarks
Structure your next phase
Zerologic builds and optimises commerce funnels — including the operational causes behind the drop-offs that appear on the checkout page.
Talk to us: partners@zerologic.io · zerologic.io
