Across the 220 lead-gen A/B test results in LiftArchive, 207 (94.1%) were reported as wins, 7 as losses and 6 as flat.
Of the 170 wins that report a percentage lift, the median is about +41%, and form tests specifically land at a median of +35.8% (or +24.5% when you count only clean A/B tests).
Treat those as the ceiling of what published tests show, not what your next test will do: public case studies skew heavily toward wins.
This page breaks the 220 results down by evidence type, page and element, then goes deep on form test results, the most-tested part of any lead-gen funnel.
Every number below was counted from our test database as it stood in November 2026.
How we counted
We took every result tagged with the lead-gen vertical in our database: n = 220, credited to 97 different teams, agencies and in-house programs.
Each result keeps the headline number exactly as the source reported it.
To compute lifts, we used results where the headline is a percentage increase (for example "+31.23% form submissions") or a before-and-after rate (for example "1.77% to 4.33% conversion", which we converted to a relative lift of +144.6%).
Results reported as "Winner", "more leads" or "2x" without a percentage were counted as wins but left out of the lift math. That leaves 170 quantified wins.
Three caveats shape every number on this page:
- Publication bias. Agencies and vendors publish wins. Ronny Kohavi and colleagues at Microsoft reported that only about one third of ideas improved the metrics they were designed to improve. Our 94.1% win rate says far more about what gets published than about how often lead-gen tests work.
- Mixed metrics. Some tests count submitted leads, others count clicks, opens or quote starts. A click lift is not a lead lift.
- Mixed rigor. Some results are single-change A/B tests, others are full redesigns or before-and-after comparisons. We split them out below because they behave very differently.
For the cross-vertical picture, see our research on A/B test win rates and average A/B test lift.
Lead-gen results at a glance
Lead-gen has the highest reported win rate of the four verticals in the database.
That is almost certainly because more of its results come from agency case studies, where wins are the whole point of publishing.
| Vertical | Results | Won | Win share |
|---|---|---|---|
| Lead-gen | 220 | 207 | 94.1% |
| SaaS | 205 | 172 | 83.9% |
| Ecommerce | 1,219 | 998 | 81.9% |
| Mobile apps | 80 | 70 | 87.5% |
The spread of lead-gen lifts is wide. Of the 170 quantified wins:
| Reported lift | Wins | Share |
|---|---|---|
| Under +10% | 21 | 12% |
| +10% to +25% | 33 | 19% |
| +25% to +50% | 37 | 22% |
| +50% to +100% | 31 | 18% |
| +100% or more | 48 | 28% |
The middle half of results sits between about +19% and +121%.
More than a quarter claim to have doubled conversions, which is a strong sign that many of these tests started from a weak baseline page.
Single-change tests report smaller lifts than redesigns
How the test was run matters more than any other factor we looked at.
| Evidence type | Results | Won | Quantified wins | Median lift |
|---|---|---|---|---|
| A/B test (one change or a tight set) | 130 | 121 | 95 | +32.0% |
| Bundled redesign | 76 | 74 | 67 | +64.8% |
| Before-and-after | 7 | 7 | 6 | +73.5% |
| Multivariate | 7 | 5 | 2 | too few |
Bundled redesigns double the median lift, but they cannot tell you which change did the work.
Before-and-after comparisons are weaker still, since traffic mix and seasonality change between periods.
One source dominates the bundled group. KlientBoost contributed 42 paid-traffic landing page rebuilds, all reported as wins with a median lift of +73% across the 39 that give a number.
Without KlientBoost, the remaining 178 lead-gen results have a median lift of +36%.
By page type
| Page | Results | Won | Quantified wins | Median lift |
|---|---|---|---|---|
| Landing page | 94 | 90 | 78 | +66.0% |
| Lead form | 43 | 40 | 35 | +31.2% |
| Homepage | 23 | 23 | 22 | +46.4% |
| 18 | 15 | 8 | +37.9% | |
| Site-wide | 17 | 17 | 13 | +33.1% |
Landing pages show the biggest numbers because most of them are full rebuilds for paid traffic.
Lead form tests are smaller but cleaner: most change one thing on a form people have already decided to fill in.
More on that page type at lead-gen landing page tests.
By element
We grouped element tags into broad families. A test can sit in more than one family.
| Element family | Results | Median lift (all) | Median lift (A/B tests only) |
|---|---|---|---|
| Forms (fields, steps, placement, progress, design) | 52 | +35.8% (n=43) | +24.5% (n=23) |
| CTA copy, placement and design | 47 | +60.0% (n=35) | +43.0% (n=23) |
| Social proof and trust | 25 | +35.0% (n=18) | +40.5% (n=10) |
| Headlines | 20 | +38.0% (n=15) | +16.9% (n=6) |
| Personalization and message match | 15 | +15.3% (n=11) | +15.3% (n=7) |
CTA results look large partly because many measure clicks rather than leads.
Across all quantified wins, the 22 results reported on a click, open or engagement metric have a median of +69.4%, against +43.1% for the 130 reported on leads, submissions, sign-ups or conversions.
Restricted to clean A/B tests on a lead metric (n=70), the median drops to +31.1%.
A clear example: The Global Leadership Summit's localized event page drove +32% clicks but no registration lift. Measure the lead, not the click.
Form test results
Forms are where lead-gen tests are most consistent.
Of 52 form tests, 51 were reported as wins, and the clean A/B tests cluster between about +16% and +53%.
The ideas that recur fall into four groups.
Multi-step forms
Splitting a long form into steps is the most repeated large win in the lead-gen data.
It works best when the first step asks an easy, low-commitment question.
| Test | Who ran it | Result |
|---|---|---|
| One-page inquiry form split into seven steps | HomEquity Bank | +154% conversion rate |
| Seven-step form cut to four steps | HomEquity Bank | +53% conversions |
| Multi-step life insurance quote form | Reassured | +31.23% form submissions |
| Long claim form split into four stages | Apexure | 1.77% to 4.33% conversion |
| Multi-step consulting enquiry form | Venture Harbour | 0.96% to 8.1% conversion (before-and-after) |
The two HomEquity Bank tests belong together. Going from one page to seven steps won, and then cutting seven steps to four won again.
More steps is not the goal; easier steps are.
The Venture Harbour figure is a before-and-after comparison, so read it as a direction, not a size. See all multi-step form tests.
Form fields
Field tests are where "fewer fields always wins" breaks down.
Removing friction usually helped, but one team won by adding a required field.
| Test | Who ran it | Result |
|---|---|---|
| 'Don't remember passport details?' option | Tinkoff Bank | +35.8% applications |
| Deal form cut to 3 required fields, rest hidden | InsightWhale | +26% conversions |
| Removing repeated form fields | VWO | +19.31% completed quote requests |
| Simpler date-of-birth field | Vandebron | +16.3% sign-ups |
| Making the profession field required | VWO | +12.52% completed requests |
| Phone-only popup form for loan leads | Mercantil Bank | +5.5% submissions, -23.6% CAC |
| Guide choice plus one extra form field | MarketingSherpa | significant drop in leads |
The pattern is about effort, not field count.
Fields that are hard to answer (passport numbers, awkward date formats, repeated questions) cost leads.
A field that helps route the visitor, like Qualicorp's profession field, can raise completions.
Browse the full set under form field tests.
Progress indicators
Once a form has steps, showing progress tends to help.
HomEquity Bank's progress tracker lifted conversions by +35%, and Tinkoff Bank's progress bar lifted credit card application submissions by +12.8%.
Form placement and visibility
Where the form sits can matter as much as what it asks.
Warner Music Group moved the sign-up form on artist sites and sign-ups went from 6.2% to 18.4%.
Hiding the form can work too: Northmill tucked a loan form behind an 'Apply' button for +7% applications.
For a step-by-step approach to running these tests, see our guide to lead-gen form optimization.
The losses and flat results
Only 13 of 220 results were reported as losses or flat, so treat each one as valuable.
They show which ideas are not safe defaults.
- Uncommon Knowledge's modern opt-in page redesign lost: the original got 19.55% more sign-ups.
- NextAfter's added value copy on a donation page cost -28% donations.
- A rounded start box on an insurance quote page (MarketingExperiments) dropped CTR from 17.79% to 17.01%.
- Booking.com moved registration steps higher on its partner join page and saw no lift.
- Letting visitors pick one of three free guides produced no change in conversions.
More examples across all verticals are in our losing A/B tests research.
How to use these numbers
Use the medians to rank ideas, not to forecast.
If a clean A/B form test in our data has a median of +24.5%, that is the median of tests someone chose to publish, not the expected result of your test.
- Start with form friction. Hard fields, repeated questions and one long page are the most consistent wins in lead-gen data.
- Measure submitted leads, and ideally lead quality. Click-based wins run larger and often don't carry through.
- Don't expect redesign-sized lifts from small changes. The +100% results are mostly rebuilds of weak pages.
- Plan for losses. In an honest program, most tests will not win.
For comparison with subscription businesses, see our SaaS A/B test results.
All of our research lives at LiftArchive research, and every lead-gen test is browsable from the lead-gen hub.