What to A/B test on a landing page, and in what order
Landing page A/B testing in the order that moves revenue: offer, headline, proof, call to action, then design, and the traffic each test needs.
Updated 28 September 2026 · 6 min read
Test in order of how much money a change can move: the offer and who the page is for first, then the headline and hero, then proof and objections, then the call to action and form, and layout and design last. Most sites can only detect big lifts, so start with big changes. At 3,000 visitors a week and a 2.5% conversion rate, a four-week test can only reliably detect a lift of about 32%.
Test a landing page in order of how much money each change can move. Start with the offer and who the page is for, then the headline and the first screen, then proof and the answers to objections, then the call to action and form, and leave layout and design details for last. The early items change whether a visitor wants what you sell. The late items change how easily they act on it, which usually matters less.
The order also follows from traffic. Most landing pages don't get enough visitors to detect small lifts, so a test only pays off if the change is big enough to measure. At 3,000 visitors a week and a 2.5% conversion rate, four weeks of testing can reliably detect a lift of about 32%, and nothing smaller. Judge every test on revenue per visitor, not clicks.
Why the order matters
Most changes don't win. In a paper on rules of thumb for online experiments, Ron Kohavi and colleagues report that about a third of ideas tested at Microsoft improved the metric they targeted, and only 10 to 20% at Bing. The ones that did win at Bing usually moved key metrics by 0.1 to 1%. Bing can detect a 0.5% change because it has millions of users. A landing page with a few thousand visitors a week can't.
Here is the maths for a smaller site. A standard shortcut for sample size, at 80% power and 5% significance, is 16 × p × (1 − p) ÷ d² visitors per version, where p is the conversion rate and d is the difference you want to detect. Turn it around to find the smallest difference a given sample can detect:
d = square root of (16 × p × (1 − p) ÷ visitors per version)
At 3,000 visitors a week, a four-week test gives 6,000 visitors per version. With p = 2.5%:
- 16 × 0.025 × 0.975 = 0.39
- 0.39 ÷ 6,000 = 0.000065
- The square root is 0.0081, or 0.81 percentage points
- 0.81 ÷ 2.5 = a 32% relative lift
Run eight weeks and the smallest detectable lift falls to about 23%. A site with 30,000 visitors a week could detect about 10% in four weeks. Try your own numbers in the minimum detectable effect calculator. If your traffic is small, A/B testing with low traffic covers what else you can do.
The practical rule on a normal site is to test changes that could plausibly move revenue by 20% or more. That means offers, promises and page structure, not shades of blue.
Before you test anything
Fix what is broken. A slow page, a form that fails on mobile or a headline that doesn't match the ad are bugs, not hypotheses. Speed in particular is worth fixing without a test. In the same Kohavi paper, a Bing experiment found every 100 milliseconds of speedup improved revenue by 0.6%.
The order to test in
| Step | What to test | Why here | Main metric |
|---|---|---|---|
| 1 | Offer and audience | Decides whether anyone wants it | Revenue per visitor |
| 2 | Headline and first screen | Decides whether they read on | Revenue per visitor |
| 3 | Proof and objections | Decides whether they believe it | Revenue per visitor |
| 4 | Call to action and form | Decides how easily they act | Revenue per visitor, form completion |
| 5 | Layout, length and design | Fine-tunes the above | Revenue per visitor |
1. Offer and audience
The offer is what the visitor gets and on what terms: the product, the price, the plan they land on, a trial or guarantee, and any bundle. The audience is who the page speaks to. A page for "small businesses" and a page for "Shopify stores doing $1M a year" can sell the same product and convert very differently.
Tests to try:
- A trial vs a money-back guarantee. See free trial vs no free trial.
- One page for everyone vs a page for your best customer segment.
- The plan the page sends people to, such as starter vs a mid-tier plan.
- A bundle or starter kit vs a single product.
2. Headline and first screen
The headline states the promise. The first screen, meaning everything visible before scrolling, decides whether people scroll. Test different promises, not word swaps: a speed promise vs a money promise, or a pain point vs an outcome. The headline guide goes deeper.
Tests to try:
- Outcome headline vs problem headline.
- Product screenshot or demo video vs a photo of the result.
- Price or "from $X" shown on the first screen vs lower down.
3. Proof and objections
Once people want it, they need to believe it. Proof is real customer reviews, real numbers and real logos. Objections are the reasons people don't buy: price, setup time, switching cost, trust. Answer them on the page.
Tests to try:
- Reviews near the top vs near the call to action.
- A short FAQ that answers the top three support questions vs none.
- A comparison table against the usual alternative vs none.
Never test made-up reviews, numbers or logos. Outtest's Builder agent follows the same rule. It writes new versions of a page and never invents claims, discounts or reviews.
4. Call to action and form
The call to action is what you ask people to do. Test what the button says and what happens after the click before where it sits or what color it is. The CTA button guide covers this in detail.
Tests to try:
- "Start free trial" vs "See pricing" as the main action.
- A form with 3 fields vs 7.
- A sticky call to action on mobile vs none.
5. Layout, length and design
Long page vs short page, one column vs two, section order and visual style. These can matter, but their effects are usually smaller and harder to detect. Test them once the earlier steps have settled.
Paid traffic needs its own pages
If most visitors come from ads, the page should repeat the ad's promise in its headline. A visitor who clicked "Invoicing for plumbers" and lands on "All-in-one business software" has to work out whether they're in the right place. Test a page built for your top ad against your standard page. The Meta ads testing guide covers the ad side.
Pick tests by what they could earn
A rough way to rank ideas is traffic × revenue per visitor × the lift you think is plausible. Take a page with 12,000 visitors a month and revenue per visitor of $2.00, so $24,000 a month:
- A new offer that might lift revenue 20% is worth up to $4,800 a month.
- A new headline that might lift it 10% is worth up to $2,400.
- A button color that might lift it 1% is worth up to $240, and at this traffic you'd never be able to prove it.
The split test ROI calculator does this for any lift. Outtest's Analyst agent does a version of the same thing automatically. It reads your analytics and payment data to find where the funnel loses the most money, and the Lead agent plans tests there first.
Tests to try, and what to measure
- Page for your top segment vs generic page. Measure revenue per visitor from that segment's traffic.
- Trial vs money-back guarantee. Measure revenue per visitor after the trial window.
- Outcome headline vs problem headline. Measure revenue per visitor.
- Demo video vs static screenshot on the first screen. Measure revenue per visitor and scroll depth.
- Price shown on the landing page vs only on the pricing page. Measure revenue per visitor.
- Reviews moved above the fold. Measure revenue per visitor.
- FAQ answering the top objections vs none. Measure revenue per visitor.
- Short form vs long form. Measure revenue per visitor and form completion.
- Long page vs short page. Measure revenue per visitor.
- Ad-matched page vs standard page for paid traffic. Measure revenue per visitor on paid traffic.
For more ideas filtered by page type, the A/B test ideas list has over 100. If you're not sure where to start at all, read what to A/B test first, and see revenue per visitor for why every test above uses it.
Questions people ask
What should I A/B test first on a landing page?+
The offer and the promise: what you're selling, to whom, at what price, with what trial or guarantee, and the headline that states it. These change whether a visitor wants the thing at all, so they have the most room to move revenue. Button colors and small copy edits rarely move it enough to measure.
How many things should I change in one landing page test?+
As many as it takes to test one clear idea. A new offer might need a new headline, hero and proof section together. You won't know which part did the work, but on normal traffic a big combined change is the only kind you can measure. Break it apart later if it wins.
How much traffic does a landing page test need?+
It depends on your conversion rate and the lift you want to detect. At a 2.5% conversion rate, detecting a 10% lift takes about 62,000 visitors per version, while a 30% lift takes about 7,000. Use a sample size calculator with your own numbers.
What metric should a landing page test use?+
Revenue per visitor, from your payment tool, whenever the page leads to a purchase. Signups and clicks are useful context, but a page that attracts more low-intent signups can win on those and lose on revenue.
Read next
Let Outtest run your split tests
AI agents read your analytics and payments, find where you lose the most money, build the fix and test it. Every test is judged on revenue, not clicks. Plans from $29 a month.