Most testing programs pick what to test based on what a stakeholder noticed last week, not on where a test is statistically likely to win. That is a resourcing problem as much as a strategy problem, because every test consumes traffic, time and a slot on the roadmap, and not every page in the funnel returns the same odds. Recent test-outcome data across thousands of ecommerce experiments makes the pattern clear enough to actually plan around, and it does not match where most teams currently spend their testing effort.
Cart and basket-page tests win most often, around 24 to 25 percent of the time, followed by checkout's first step near 21 percent and product pages around 17 percent. Checkout carries the widest spread between wins and losses of any stage, and mobile tests outperform desktop tests by roughly 6 to 7 percentage points. Prioritize cart and mobile before chasing homepage redesigns.
Where do A/B tests actually win in 2026?
Pooled results from large-scale ecommerce testing programs show a consistent ranking by page type. Basket and cart pages post the highest win rates, in the mid-20s percent range, because the visitor has already committed to buying and the page has a narrow job: remove friction between "I want this" and "I paid for this." Checkout's first step sits close behind at roughly 21 percent. Product detail pages trail at around 17 percent, largely because PDP tests compete against a harder problem, convincing someone to want the product at all, which is a bigger lift than removing friction from a decision already made. Homepage and top-of-funnel tests, while common on testing roadmaps, tend to show the lowest and least reliable win rates of any page type, because the changes being tested are furthest from the purchase decision and easiest to drown in noise.
Why does checkout behave differently from every other stage?
Checkout is the one funnel stage where test outcomes are genuinely bimodal rather than clustered near a modest average. Recent data puts checkout around a 36 percent win rate against a 29 percent loss rate, the widest gap between wins and losses of any stage measured. That split makes sense once you think about what checkout actually is: the last page before money changes hands, with almost zero tolerance for added friction. A change that removes a real blocker, an unnecessary field, a hidden guest checkout option, an unclear shipping cost, wins decisively because it directly unblocks a purchase that was about to happen anyway. A change that adds even minor confusion, a reordered field, an unfamiliar button label, loses decisively for the same reason. Checkout tests are not for the risk-averse, but they are where a single good idea pays for a quarter of testing effort.
| Funnel stage | Approx. win rate | Why |
|---|---|---|
| Cart / basket | ~24-25% | Visitor already committed, narrow job to do |
| Checkout, step 1 | ~21% | Close to purchase, moderate friction tolerance |
| Product detail page | ~17% | Has to build desire, not just remove friction |
| Checkout, full flow | ~36% win / ~29% loss | Bimodal: fixes win big, added friction loses big |
| Homepage / top of funnel | Lowest, least reliable | Furthest from purchase decision, high noise |
What does Baymard's checkout research add to this picture?
Baymard Institute's long-running checkout usability benchmarking, built on detailed guideline audits of major ecommerce sites, has repeatedly found dozens of specific, fixable usability issues on the average checkout when measured against its full guideline set. That number matters for prioritization: it means most checkouts are not one field away from optimal, they are carrying a backlog of small, compounding friction points, and a structured audit against known usability guidelines will surface more real test candidates than guessing. One issue Baymard has flagged repeatedly is guest checkout visibility on mobile, where the option gets placed low enough on the page to be obscured by the on-screen keyboard, a purely positional bug that costs completed orders for no design reason at all.
Run a structured checkout audit against a known guideline set before you run a single test. Guessing at checkout problems from analytics drop-off alone tells you where people leave, not why, and a proper usability pass usually surfaces more testable ideas in a day than a quarter of ad hoc hypotheses.
Why does mobile outperform desktop as a testing target?
Mobile-only tests have shown roughly a 6 to 7 percentage point higher win rate than desktop-only tests in recent pooled data. Two things explain the gap. First, mobile experiences on most sites simply have more unaddressed problems, since design and QA effort still skews toward desktop on many teams, which leaves more low-hanging fruit. Second, mobile now carries the majority of ecommerce traffic on most sites while converting at close to half the desktop rate, which is exactly the kind of gap our own analysis of the mobile conversion gap covers in more depth. More traffic and more unfixed problems together mean mobile tests reach significance faster and win more often, which makes mobile the higher-ROI place to point a limited testing budget in 2026.
How should a small team actually allocate its testing roadmap?
- Put cart and checkout first. They have the highest win rates and sit closest to revenue, so a win there shows up in the topline fastest.
- Run a structured checkout usability audit before writing test hypotheses, rather than guessing from a drop-off report alone.
- Default new tests to mobile unless you have a specific desktop-only hypothesis. The traffic and the upside both favor it.
- Treat product page and landing page tests as valuable but slower burning. Budget for them, but do not expect the same hit rate as cart and checkout.
- Deprioritize homepage redesigns as a testing target. They are the hardest page to move with the least reliable signal, better suited to periodic strategic review than a testing queue.
This is also why so few marketers actually run a consistent program: credible estimates put active landing page testing at under one in five marketers, not because testing doesn't work but because low-traffic pages take too long to reach significance and testing tools take setup time a stretched team rarely protects. Fixing that is less about tooling and more about sequencing, start where wins come fastest, and use the credibility from early wins to justify the slower-burning tests later.
On accounts with limited traffic, we run cart and checkout tests first specifically to build a track record of wins before asking a client to be patient with a six-week homepage test. Sequencing tests by expected win rate, not by internal politics, is the single biggest lever for keeping a testing program funded past its first quarter. See our broader take on conversion optimization for how we structure a full testing roadmap.
Frequently asked questions
Which funnel stage has the highest A/B testing win rate?
Cart and basket page tests tend to win most often, with win rates around 24 to 25 percent in recent large-scale test analyses, ahead of checkout's first step at roughly 21 percent and product detail pages at around 17 percent. Basket-stage tests sit closer to the purchase decision with fewer variables in play, which makes clean wins easier to produce.
Why does checkout have both a high win rate and a high loss rate?
Checkout tests are polarizing because the page has little tolerance for friction. A change that removes a genuine blocker wins decisively, while a change that adds even minor confusion at the final step loses decisively. Recent test data puts checkout around a 36 percent win rate against a 29 percent loss rate, the widest spread of any funnel stage, which makes it high-reward but higher-risk to test than earlier pages.
Should small businesses A/B test mobile or desktop first?
Mobile, in most cases. Mobile-only tests have shown roughly a 6 to 7 percentage point higher win rate than desktop-only tests, and mobile traffic typically converts at close to half the desktop rate, which means there is more room to fix and more traffic to test against on most sites.
How many usability issues does a typical checkout have?
Baymard Institute's ongoing benchmarking of major e-commerce checkouts has repeatedly found dozens of specific, fixable usability issues per site when measured against its detailed guideline set, which is far more than most teams assume before they run a structured audit.
Why do so few marketers actually A/B test their landing pages?
Estimates put active landing page testing at under one in five marketers, despite testing programs showing meaningful average lift when run consistently. The gap is usually traffic volume and time, not belief in testing. Low-traffic pages take too long to reach significance, and testing tools take setup time teams without a dedicated CRO resource rarely protect.
The takeaway
Test allocation is a resourcing decision, and the data says most teams have it backwards: heavy investment in homepage and top-of-funnel tests that rarely win, light investment in cart, checkout and mobile, where wins come fastest and matter most. Reorder the roadmap around where tests actually win, not around what a stakeholder noticed last week, and a testing program earns the credibility, and the budget, to tackle the slower, harder pages later.