Review a page
Landing page analyzer
A free review of your landing page, scored on eight checks.
A/B testing
Landing page A/B testing: how to split test a landing page
Landing page A/B testing shows half your visitors the current page and half a changed version, then counts which one turns more of them into leads or sales. This guide covers what to test, how much traffic a test needs, how to split paid traffic, and how to read the result.
What is landing page A/B testing?
Landing page A/B testing is a controlled experiment on a page built for one campaign and one action. Visitors are split at random between the current page, the control, and a changed version, the variant, at the same time. The version that turns more of them into leads, sign-ups or sales wins, once the sample you planned before launch has been reached.
Landing pages are often the easiest place to start testing. The traffic source is known, there is a single goal, and the page usually gets paid traffic, so every point of conversion rate changes what each lead or sale costs. People also call it landing page split testing; the method is the same.
Types of landing page testing
Landing page testing covers four kinds of test: an A/B test on the same URL, a split URL test between two addresses, a multivariate test of several elements at once, and checks before launch that use no live traffic. The first three measure real visitors; the fourth finds problems before you spend traffic on them.
| Method | What changes | Best for | Traffic needed |
|---|---|---|---|
| A/B test | One element or section, same URL | Headline, form, call to action, proof, offer | Depends on baseline and lift; set by a sample size calculation |
| Split URL test | The whole page, each version at its own address | A redesign, a new layout, a page built in another tool | Same rules as an A/B test |
| Multivariate test | Several elements, every combination | Finding how elements work together on busy pages | Much more, because traffic is split across many versions |
| Checks before launch | Nothing goes live | User tests, copy feedback, a page review, a tracking check | None |
For the full list of test types, see CRO testing. A free review before launch is what the landing page analyzer does: it reads the page and scores eight checks, with no test traffic needed.
Why A/B test a landing page?
A/B testing a landing page tells you whether a change raised conversions or only looked better in a meeting. Opinions about headlines, images and form length are cheap and often wrong. A test measures what real visitors from your real traffic source do, so you keep changes that work and drop the ones that do not.
- Lower cost per lead or sale. If the same ad spend sends the same visitors to a page that converts more of them, each conversion costs less.
- Evidence for a redesign. A split URL test shows whether a new page beats the old one before you replace it for everyone.
- Learning that carries over. A result about what your visitors need to see before they act, such as a price or a guarantee, often applies to your other pages too.
- Protection against harm. A change that would have lowered conversions is caught on half the traffic for a few weeks, not on all of it for months.
How to A/B test a landing page in 8 steps
To A/B test a landing page, choose one page and one goal, find the problem in your data, write a hypothesis, calculate the sample size, build and check the variant, split the traffic at random, wait for the planned sample, then read the result and roll out the winner. The steps below cover each one.
-
Choose the page and the goal
Pick a landing page with steady traffic and one conversion that matters: a form submission, a booked call, a sign-up or an order. Write down the primary metric that will decide the test, such as completed forms per visitor, before anything else.
-
Find the problem in your data
Look at where visitors leave and why. Analytics shows drop off by device and traffic source, heatmaps and session recordings show what people do on the page, form analytics shows which field stops them, and a short poll asks what was missing.
-
Write a hypothesis
State what you saw, what you will change, for whom, and what you expect to happen: because we saw [evidence], we believe [change] for [visitors] will [result], measured by [metric]. A hypothesis makes a losing test useful, because it tells you which belief was wrong.
-
Calculate the sample size
Enter the page's current conversion rate and the smallest lift worth detecting into the A/B test sample size calculator. Divide the total by daily visitors to get the days, and round up to whole weeks, as set out in how long to run an A/B test.
-
Build the variant and check it
Make the change, then check both versions on phones, tablets and desktops in the main browsers. Submit the form on each version and confirm the conversion is recorded once, with the right version attached. A variant that breaks on one browser ruins the test.
-
Split the traffic and launch
Send visitors at random, usually 50/50, and make sure each person keeps seeing the same version. Keep the ads, keywords, audiences, emails and budget that feed the page the same for the whole test, and do not edit either version while it runs.
-
Wait for the planned sample
Leave the test alone until each version reaches the sample you calculated, in whole weeks. Check only that it is running correctly. Stopping the first time one version looks ahead is the most common way to declare a winner that is really noise.
-
Read the result and act on it
Check the primary metric with a statistical significance calculator, then look at devices and traffic sources separately. Roll out a clear winner, keep the control after a loss, and record the result, the numbers and what you learned either way.
What to test on a landing page
The landing page elements most worth A/B testing are the ones your data points to: the headline and first screen, the offer, the call to action, the form, the proof near the decision, the page length and the distractions around the goal. The table pairs each element with tests to try and the evidence that suggests it.
| Element | Tests to try | Evidence that points here |
|---|---|---|
| Headline and first screen | Repeat the promise of the ad or email; name the outcome, not the product | Most visitors leave without scrolling |
| Offer | A different lead magnet, a free trial against a demo, a clearer price | Visitors scroll the whole page but few act |
| Call to action | Button wording that says what happens next, its position, a sticky button on mobile | Few clicks on the button, or clicks only far down the page |
| Form | Fewer or optional fields, steps instead of one long form, a form in a popup | Many form starts, few submissions; one field where people stop |
| Proof | Reviews, client names or a guarantee placed beside the form or button | Polls or sales calls show doubt about trust or results |
| Page length | A short page against a long page that answers the main questions | Different behavior from warm and cold traffic sources |
| Distractions | Remove the navigation and links that lead away from the goal | Clicks on the menu or footer links instead of the call to action |
| Images and video | A product in use against a stock photo; a short video against none | Recordings show visitors skipping past the hero image |
Before testing a headline, run both versions through the free headline analyzer, so neither loses to a flaw you could have fixed first. For more ideas and why a page underperforms, see landing page optimization and landing page not converting.
How much traffic do you need to A/B test a landing page?
The traffic a landing page A/B test needs depends on two numbers you know before launch: the page's current conversion rate and the smallest lift worth detecting. There is no single minimum. A page that converts rarely, or a change expected to make a small difference, needs far more visitors than a big change on a page that converts often.
Say a landing page converts at 5% (an example input, not a typical rate: use your own) and you want to detect a 25% relative lift, from 5% to 6.25%, at 95% confidence and 80% power, with one variant against the control. 400 visitors a day reach the page.
- Sample per version: 5,333 visitors.
- Total sample: 5,333 x 2 versions = 10,666 visitors.
- Days: 10,666 / 400 a day, rounded up to 27 days, so run the test for 4 weeks.
| Smallest lift to detect | Visitors per version | 100 visitors a day | 500 visitors a day | 2,000 visitors a day |
|---|---|---|---|---|
| 15% (5% to 5.75%) | 14,193 | 41 weeks: change the plan | 9 weeks: change the plan | 3 weeks |
| 25% (5% to 6.25%) | 5,333 | 16 weeks: change the plan | 4 weeks | 1 week |
| 50% (5% to 7.5%) | 1,471 | 5 weeks | 1 week | 1 week |
Every figure comes from the same formula as the sample size calculator, with a one-week floor; two full weeks is safer. Anything over 8 weeks is marked "change the plan". On a low traffic landing page that usually means testing a bigger change, measuring a step nearer the change such as form starts, or fixing the page without a test.
How to split test landing pages with paid traffic
To split test landing pages with paid traffic, send the ad to one URL and let a testing tool divide visitors at random, or use the ad platform's own experiment feature to send part of the traffic to each page. Either way, keep the ads, keywords, audiences and budget fixed until the test ends.
Why the traffic source must stay the same
A landing page converts differently for different visitors. If a new campaign, a budget change or a new audience starts halfway through, the mix of people reaching each version shifts, and the result measures the traffic change as much as the page change. Plan the test for a period with no planned campaign changes.
Do not run two ads to two pages
Pointing two separate ads at two separate pages is not an A/B test. The ad platform sends each ad to different people at different times as it optimizes delivery, so the two pages are not seen by comparable groups. The split has to be random, between visitors who clicked the same ad.
Test one traffic source at a time
Visitors from search ads, social ads and email often want different things. If a page gets traffic from several sources, either test with one source, or plan the sample so you can read each source on its own, and expect that a winner for one source may not win for the others.
Does landing page A/B testing affect SEO?
Landing page A/B testing does not hurt SEO when it follows Google's published guidance, A/B testing best practices for Search. Many paid landing pages are kept out of search on purpose, but if yours is indexed, four rules apply.
- No cloaking. Do not show Googlebot a different page from the one visitors see.
- A canonical link on the variant. In a split URL test, the variant's address carries
rel="canonical"pointing to the original page. Google recommends this rather than anoindextag. - A temporary redirect. Send visitors from the original to the variant with a 302 redirect, not a permanent 301, so the original stays indexed.
- Remove the test when it ends. Run only as long as the test needs, then update the page and take the test code and the variant URL down.
How to read a landing page A/B test result
Read a landing page A/B test result on the primary metric you chose before launch, once the planned sample is in. Check that the difference is statistically significant, then look at devices and traffic sources separately, and for lead generation pages, check the quality of the leads as well as the count.
Check significance, not the first good day
A difference between two versions can be chance. A significance check tells you how surprising the result would be if the change did nothing. The p-value calculator page explains what that number does and does not tell you.
Look at mobile and desktop separately
Landing page changes often work on one device and not the other. In Convertica's sticky CTA test, the mobile rate of order completion rose and desktop did not change, so the change went live on mobile only. A combined figure would have hidden that.
Follow the lead past the form
A shorter form or a softer offer can raise form submissions while lowering the share of leads who buy. Where you can, match each lead to what happened next in your CRM, and judge the winner on qualified leads or sales, not only on submissions.
Record every result
Write down the hypothesis, the dates, the sample, the result and what you learned, including for tests that lost or showed no difference. That record stops the same idea being tested twice and shows which beliefs about your visitors held up.
Great results. easy to work with, and do proper split-testing. Even noticed when I accidentally made a change to a page they were testing, and they stopped the test and started again to make sure I didn't distort results. The most important thing to know: they had a positive result on revenue.
Landing page A/B testing examples from Convertica case studies
These three Convertica case studies tested a change on a page built to get one action: a sticky call to action, a quote form in a popup, and a rebuilt submission process. Each figure is as stated in the original case study, with its metric and test length. Results from one site are not a forecast for another.
| Case study | What was tested | Test | Stated result |
|---|---|---|---|
| Sticky CTA | A call to action bar that stays on screen, on an online store's landing page | 20-day A/B test, 98% confidence on mobile | +252.9% rate of order completion on mobile; desktop showed no significant change |
| Home services | A quote form in a popup, opened by call to action buttons, against a form on the page | Desktop and mobile, 97% confidence, test length not published | +45.08% conversion rate (successful bookings), desktop and mobile combined |
| Empire Flippers | A rebuilt, step-by-step valuation tool submission process, split URL tested 50/50 | First test: two weeks. Result after rollout: 47 days | +40% conversions in the first test; +51.6% conversions (valuation tool submissions) in 47 days |
More tests, with what they changed and how long they ran, are on the A/B testing examples page and in the CRO case studies.
Common landing page A/B testing mistakes
The most common landing page A/B testing mistakes are stopping early, testing without enough traffic, changing the traffic source mid-test, measuring clicks instead of leads or sales, and testing a change with no evidence behind it. Each one can produce a winner that is not real, or hide one that is.
- Stopping when one version looks ahead. Results swing in the first days. Wait for the planned sample.
- Starting a test the traffic cannot finish. Do the sample size sum first; if it says months, change the test.
- Changing campaigns during the test. New ads, audiences or budgets change who reaches each version.
- Measuring the wrong thing. Button clicks are easy to raise; leads, qualified leads and sales are what pay.
- Testing guesses. A test idea without evidence is the one most likely to waste weeks of traffic.
- Ignoring devices. A change that wins on desktop can lose on mobile, and the combined figure hides it.
- Not checking the split. If one version got clearly more visitors than planned, something in the setup is wrong.
- Testing what is broken. A form that fails on one browser or a page that loads slowly should be fixed, not tested.
Landing page testing tools
Landing page testing tools come in three kinds: landing page builders and website platforms with A/B tests built in, dedicated web testing tools that work on any page, and ad platforms that can split traffic between two pages. Choose by where your page is built and how much control you need over the split and the goals.
- Built into the page builder. Many builders and marketing platforms let you make a variant and split traffic without extra code. HubSpot, for example, documents how to run an A/B test on a page. Check your own builder's help pages for what its tests can measure.
- A dedicated testing tool. Works on pages built anywhere, with more control over targeting, goals and segments. Compare them in the guide to A/B testing tools.
- The ad platform's experiments. Useful when the page only gets traffic from one platform; the platform divides the traffic before the click.
Whichever you use, run one A/A test or check the first days of a real test to confirm visitors are split as planned and each conversion is counted once.
When not to A/B test a landing page
Do not A/B test a landing page when the traffic cannot finish the test, when the problem is clear and broken, or when the change is required anyway, such as a legal notice or a price change. In those cases, make the change and watch the numbers, or use research to decide.
- The fix is obvious. A broken form, a slow page, a headline that contradicts the ad or a missing price should be fixed straight away.
- The traffic is too low. If the sample size calculator says months, test a much bigger change or rely on user tests and polls.
- The page is about to change. A test on a page that will be replaced next month answers a question nobody will ask.
Not sure which problems on your landing page are clear fixes and which need a test? Convertica's free CRO audit reviews the page for people and for AI agents and lists every issue it finds, with why it matters and the evidence. If you would rather have the fixes built, the one page rebuild has Convertica's team build every fix the audit found on that page; it builds the fixes, it does not test them.
Getting help with landing page testing
If you want a team to plan and run landing page A/B tests for you, Convertica quotes A/B testing one by one, as a custom proposal: see A/B testing services to ask for an A/B testing proposal. For testing on other kinds of page, see the guides to ecommerce A/B testing, Shopify A/B testing and price testing.
Free tools to plan and read a landing page test
Review the page, plan the sample and read the result with Convertica's free tools. Each calculator shows its formula and a worked example.
Plan a test
A/B test sample size calculator
Visitors each version needs, and how many days that takes at your traffic.
Check a result
Statistical significance calculator
Whether a finished A/B test result is bigger than chance.
A/B testing services
Want your A/B tests run for you?
Convertica runs A/B tests for clients, quoted one by one as a custom proposal. Monthly traffic comes first, because it decides whether a test on your site can reach a clear result, and how long it takes.
The split tests in Convertica's case studies were run by its team for past clients, and each result is published with its metric, timeframe and source.
Short answer
Landing page A/B testing splits visitors at random between your current landing page and a changed version, at the same time, and measures which one gets more conversions. To run one, find a problem in your data, write a hypothesis, set the sample size before launch, keep the traffic source steady, and read the result only at the planned sample.
- Test one clear change against the control, on one goal, with the traffic source held steady for the whole test.
- Calculate the sample size before launch: a low converting page or a small expected lift can need months of traffic.
- Use a split URL test for a full redesign, and a 302 redirect with a canonical link so the test does not hurt SEO.
- Read the result by device and traffic source, and check lead quality as well as lead count.
- Fix clear problems without a test, and save tests for changes you are unsure about.
Questions about landing page A/B testing
What is landing page A/B testing?
Landing page A/B testing is a controlled experiment that shows part of your visitors the current landing page and part a changed version, at the same time, and counts which version converts more of them. Because both groups arrive in the same period from the same source, the difference can be put down to the change once the sample is big enough.
How do you A/B test a landing page?
Pick one page and one goal, find the problem in your data, write a hypothesis, calculate the sample size, build the variant and check it on every device, split traffic at random, keep the ads and audience the same, wait for the planned sample, then check significance and roll out the winner.
What is the difference between landing page testing and landing page A/B testing?
Landing page testing is the wider term: it covers A/B tests, split URL tests and multivariate tests with live traffic, and also pre-launch checks such as user tests, copy tests with a panel and an automated page review. Landing page A/B testing means one live experiment between a control and a variant.
How much traffic do you need to A/B test a landing page?
It depends on the page's current conversion rate and the smallest lift you want to detect, so there is no single number. Enter your own figures in the free A/B test sample size calculator: if the answer is months of traffic, test a bigger change or a nearer goal, or fix the page without a test.
Should I use a split URL test or an A/B test for a landing page?
Use an A/B test on the same URL for changes to parts of the page, such as the headline or the form. Use a split URL test, where each version has its own address, for a full redesign or a page built in a different tool. Both split traffic at random and are read the same way.
Can you A/B test a landing page with Google Ads traffic?
Yes. Either send the ad to one URL and let a testing tool split visitors, or use your ad platform's own experiment feature to send part of the traffic to each page. Keep the ads, keywords, audiences and budget the same while the test runs, or the traffic mix changes the result.
Does A/B testing a landing page hurt SEO?
Not if you follow Google's guidance: show search engines the same pages visitors see, put a canonical link on the variant URL that points to the original, use a temporary 302 redirect for split URL tests, and remove the test once you have an answer. Many paid landing pages are not indexed at all.
What should you test first on a landing page?
Test first where your data shows the biggest problem: the first screen if most visitors leave without scrolling, the form if many start it and few finish, the call to action if few click it. A headline that repeats the ad's promise is a common first test for paid traffic.
How long should a landing page A/B test run?
Run it until each version reaches the sample size you calculated before launch, rounded up to whole weeks, and at least one full week so every day of the week is counted. Do not stop early because one version is ahead: early results swing widely and stopping on a good day produces false winners.
Does Convertica run landing page A/B tests?
Convertica quotes A/B testing one by one, as a custom proposal, through its A/B testing services page. The free CRO audit, which you can run now, reviews a landing page and lists every issue it finds, with why it matters. The one page rebuild has Convertica's team build every fix the audit found on that page; it does not test them.
