Landing page optimization: what the evidence supports
Landing page optimization means getting more of a page’s visitors to do the one thing the page is for. Most advice about it cannot be checked. The five changes below each have a published study behind them: one main call to action, a headline a stranger understands quickly, proof placed where people see it before deciding, a page built for phones, and a page that loads fast enough. For each one we say how strong the study is, because one test on one site and a dataset of thousands of pages are not the same kind of evidence.
What number are we trying to move?
First, a baseline. In Unbounce’s data from July 2023 to July 2024, the median software (SaaS) landing page got 3.8% of its visitors to do what it asked; across all industries the median was 6.6%.[1][2] Where visitors come from moves that a lot. On SaaS pages the median was 16.9% for visitors from email, 5.1% from Google search ads, 2.9% from paid social and 0.3% from display ads.[1] The colder the traffic, the more work the page itself has to do.
| Change | What was measured | How strong |
|---|---|---|
| One main call to action | 13.5% average with one link or call to action, 10.5% with three (18,639 pages)[3] | Large sample; compares pages, not a controlled test; 2018 |
| A headline understood fast | People often leave pages in 10–20 seconds; about 57% of viewing time is above the fold[4][5] | Established eye-tracking and usage research; not a test of headlines |
| Proof seen before deciding | Three lines of testimonials: purchases up 34%; an earlier test with the same testimonials further down the page: no clear effect[6] | One site, one test |
| Built for phones | 79% of visits to SaaS landing pages were on mobile devices[1] | Large dataset; says who visits, not what lifts |
| Fast enough to load | No single figure we could trace to a study with its method | Treat as basic upkeep |
Does one call to action beat three?
On the largest sample we could check, yes. In Unbounce’s 2018 analysis of 18,639 landing pages, pages with a single link or call to action averaged 13.5%, pages with two 11.9%, and pages with three 10.5%.[3]
Every extra button is a second decision. “Start free” next to “Book a demo” next to “Watch the video” means some visitors pick none. The limit: this compares pages, it is not a controlled test, and pages with one call to action may simply be better-run pages. The cost of following it is close to zero, though. Keep one button, and turn the other actions into plain text links.
Does a clear headline beat a clever one?
Nielsen Norman Group found that people often leave web pages within 10–20 seconds,[4] and in a 2018 eye-tracking study of 120 people, about 57% of page-viewing time was spent above the fold.[5] Your headline and the first screen get most of the time a visitor gives you.
The check is cheap. Show the first screen to someone outside your company for five seconds, hide it, and ask what the product does and who it is for. If they cannot say, no button colour will fix it. Rewrite until a stranger can say it back to you.
Where should proof go on the page?
Higher on the page, where people see it before they decide. WikiJob, a careers site, added three lines of testimonials to its aptitude-tests page and purchases rose 34%, according to the case study published by VWO, the testing tool it used. An earlier test that showed the same testimonials further down the page made no clear difference.[6] The case study does not say exactly where the winning testimonials sat.
This is one site and one old test, so treat it as direction, not a rule. The idea behind it is simple: proof reduces risk, so it should be seen before the decision, not after it. Placing it right beside the button, the form or the price is our suggestion, not something this test measured.
Does building for phones first matter?
In Unbounce’s data, 79% of visits to SaaS landing pages were on mobile devices.[1] If your page is a desktop design squeezed onto a phone, most visitors may be seeing its worst version. Check your own split first; if yours differs, build for your split. In practice this means the main button is reachable without scrolling past a wall of text, the headline fits without shrinking, and the page loads on a phone connection.
How much does page speed matter?
We cannot give you one clean figure, and we would rather say so than repeat one. The speed figures we found in circulation could not be traced to a study with its method, so we do not cite them. What holds without a number: a page that is slow for real visitors costs you some of the seconds above, and it makes every other problem worse. Fix what is clearly slow, then move on to changes with better evidence.
This is one thing Convertlyft measures from your own visitors: how fast the page loads for most visitors (LCP), how quickly it reacts to a tap (INP) and how much it jumps (CLS). See how Convertlyft works.
Which change should you make first?
Our suggested order, from cheapest and best supported to the rest:
- The headline. If a stranger cannot say what you do after five seconds, the rest matters less.
- One call to action. Free to change, with the largest sample behind it.
- Phones and speed. Basic upkeep that every other change depends on.
- Where your proof sits. Move the proof you already have higher, where people see it before they decide, before collecting more. Right beside the button is worth trying, but untested.
Whatever you change, measure your own page before and after. Someone else’s case study tells you what to try; your own numbers tell you whether it worked. For the benchmarks in more detail, see the benchmarks post, with every source listed.
Questions
What share of visitors should a landing page turn into sign-ups?
In Unbounce’s data from July 2023 to July 2024, the median software (SaaS) landing page got 3.8% of its visitors to do what it asked, and the median page across all industries 6.6%. Compare your page with the figure for its own kind of traffic.
How many calls to action should a landing page have?
One main action. In Unbounce’s 2018 analysis of 18,639 pages, pages with one link or call to action averaged 13.5%, against 10.5% for pages with three. That compares pages, it is not a controlled test.
Where should testimonials go on a landing page?
Higher on the page, where people see it before deciding. In one published test, WikiJob added three lines of testimonials and purchases rose 34%; an earlier test with the same testimonials further down the page showed no clear effect. The test does not say exactly where the winning testimonials sat, and it is one site and one test.
How do I know which of these my page is failing?
Measure it. A funnel shows the step where people leave, recordings show what they did on the page, and a heatmap shows where they clicked and how far they scrolled.
Sources
- Unbounce, SaaS benchmarks, from its 2024 benchmark report: 3.8% median; medians by source of 16.9% (email), 5.1% (Google search ads), 2.9% (paid social), 0.3% (display); 79% of SaaS landing-page visits on mobile. Data window 23 July 2023 to 23 July 2024 (methodology). Pages checked 23 October 2026.
- Unbounce, 2024 benchmark report: 6.6% median across all industries, more than 41,000 landing pages. Page checked 23 October 2026.
- MarketingProfs, report on Unbounce’s landing page analysis, published 24 April 2018: 18,639 pages; average 13.5% with one link or call to action, 11.9% with two, 10.5% with three. Page checked 23 October 2026.
- Nielsen Norman Group, article on how long people stay on web pages, published 11 September 2011: people often leave web pages in 10–20 seconds. Page checked 23 October 2026.
- Nielsen Norman Group, Scrolling and attention, published 15 April 2018: about 57% of page-viewing time above the fold; 120 participants, more than 130,000 eye fixations. Page checked 23 October 2026.
- VWO (Wingify), WikiJob case study, dated 11 April 2011 in the page’s data: three lines of testimonials, purchases up 34%; the same testimonials further down the page in an earlier test, no discernible impact. Page checked 24 October 2026.