They win 10% of the time — and they carry the worst measurement trap in testing. A split URL test routes traffic to a genuinely separate page rather than altering the original. Win rate is 10%, but the median winner moves +32.3%, among the larger effects in this data, because you are testing a whole page rather than one element.
Big swings, modest hit rate, and a setup mistake that manufactures fake wins. Read the warning below before you trust a split URL result.
| Tests analysed | 1,000+ |
| Beat control | 10% |
| No measurable difference | 73% |
| Lost to control | 18% |
| Median lift when it won | +32.3% |
Typical winning range: +17.6% to +57.5% for the middle half of winners. Median traffic per variant was 996 visitors. Most ran on landing pages (318), product pages (244), homepages (242).
Traffic is split between two addresses. The variant is a different page at its own URL, not a modified version of the control.
Adjacent categories, and how often they win: layout (15%), hero image (12%), styling (11%).
| page type | win rate | tests |
|---|---|---|
| landing pages | 6% | 250+ |
| product pages | 9% | 100+ |
| homepages | 13% | 100+ |
| lead capture pages | 10% | 100+ |
| content pages | 8% | 50+ |
| pricing pages | 14% | 36 tests |
Only page types with at least 30 split URL tests appear here.
A split URL variant is a whole separate page, so every description of what changed reduces to "traffic goes to a different page". The detail lives on the variant page itself, which belongs to the customer.
The trap. If your conversion goal is a pageview of the variant's own URL, every visitor sent to the variant converts instantly upon arrival, while the control converts almost never. That produces a variant at ~100% against a control near 0%, and a lift in the thousands of percent. It is not a win; it is the goal measuring the redirect. We found and excluded these when building this benchmark — the worst examples record lifts in the tens of thousands of percent. If a split URL test is showing you an implausibly large win, check whether the goal fires on the variant page itself before you ship anything.
| change type | win rate | tests |
|---|---|---|
| CTA colour | 16% | 32 |
| Layout | 15% | 351 |
| Form | 12% | 81 |
| CTA copy | 12% | 315 |
| Hero image | 12% | 160 |
| Price framing | 11% | 117 |
| Styling | 11% | 1414 |
| Split URL (this page) | 10% | 1165 |
| Headline | 8% | 918 |
| Social proof | 8% | 196 |
| Body copy | 6% | 337 |
A win rate only means as much as the thing being counted, so this is what these tests were actually optimising for.
| measured outcome | share of tests |
|---|---|
| Purchase / order complete | 46% |
| Lead / form submission | 16% |
| Signup / registration | 8% |
| Booking / appointment | 7% |
Outcomes below 5% of the set are not listed, so these do not sum to 100. How the outcome was measured moves the win rate a lot — see the methodology.
These rates are not like-for-like. How often a test wins depends partly on what it chose to measure: a goal that records a soft signal — a click, a scroll — clears the bar more often than one that records a purchase. Across this data, tests measured against a proxy goal win 13%, against 9% for tests measured against a business outcome.
Categories differ a lot in that mix — proxy goals account for anywhere from 2% to 41% of a category's tests, and split URL tests sit at 20%.
Hold the metric fixed and the spread narrows sharply. Comparing only tests measured against a business outcome, split URL wins 10% (8–13%) rather than 10%, and the gap between the highest and lowest category falls from 9 points to 6 — at which point the intervals overlap and the categories are not statistically distinguishable.
So read the table as a description of what happened, not as a ranking of what works. If you want the comparison to mean something, compare categories that were measured the same way. The methodology page has the full breakdown.
When the variant is too different to build by modifying the original — a new template, a different flow. For a single element, an on-page test is easier to read.
Whole-page variants are high-variance. More of them miss, and the ones that land move more, because more changed.
The conversion goal must sit past the variant, not on it. If control and variant cannot both reach the goal page, the test is measuring the redirect.
This is every Mida account that ran a readable test — in-house marketers, founders, product teams, and agencies working on client sites. Nothing here is filtered by who ran the experiment or how experienced they are.
Low win rates are normal in experimentation, including at the top end. Microsoft's experimentation team, reporting on its own platform, found that only about one third of ideas improve the metric they were designed to improve — and that roughly another third actively hurt it. That is a dedicated experimentation organisation with research, prioritisation and review behind every test.
A mixed population like this one runs below that. The gap is roughly what disciplined practice buys you: ideas grounded in research rather than opinion, one variable at a time, and tests built so the result can actually be read.
So treat these as a general base rate, not a target. If you would rather not close that gap the slow way, find a CRO agency in the United States, the United Kingdom, the Netherlands or Australia — or browse every region.
A win is a variant that beat its control on that test's primary goal with a statistically significant result. Tests that never got enough traffic to say anything either way are excluded. The methodology has the full detail, including what these numbers cannot tell you.
Mida.so is a super lightweight A/B testing tool to help you experiment, analyze and implement conversion strategies in minutes.
Still deciding?
A quick screen share on your actual site — no slides, no generic tour. Just your questions answered.
30 min · no commitment · no sales pressure