What makes a landing page convert: 8 factors and 4 killers

A landing page converts when someone arriving from one specific source finds what that source promised and can act on it without stopping to think. The median landing page does that 6.6% of the time. In SaaS the median drops to 3.8%, while the top quartile sits at 11.6%, so the distance between an average page and a good one is roughly three times the result on the same traffic.

Those numbers come from the Unbounce Conversion Benchmark Report, built on 57 million conversions across 41,000 landing pages. The distance they describe is almost never an aesthetic distance. The factors below are ordered by how much they move the number, not by how satisfying they are to work on.

What we measured

Four numbers decide whether any change is worth making. Collect them before touching the page.

• Conversion rate per traffic source, never blended. A page at 5% overall is often 12% on email and 1.4% on paid social, and the average hides both facts.
• Largest Contentful Paint and Interaction to Next Paint from field data on real phones, not a lab run on a laptop.
• Form start rate and form completion rate, counted separately. If nobody starts the form, the offer is the problem. If they start and abandon, the form is.
• Scroll depth to the primary call to action. If two sessions in three never reach the button, nothing below it matters yet.

The 8 factors that decide it

1. Message match with the source

The visitor clicked a specific promise. If the headline does not repeat that promise in roughly the same words, the first two seconds go to checking whether this is the right place instead of reading the offer. CXL took apart eleven pages where the wording of the ad never reappears on the page it points to (CXL). It is the cheapest fix on the list and usually the one that moves the most.

2. One page, one action

Every extra thing a visitor can do takes room from the one thing that matters. Site navigation, a chat bubble, a newsletter box, a second button saying "learn more": each one is an exit. A landing page is not a homepage. If a link does not serve the conversion, it belongs somewhere else.

3. The first screen carries most of the attention

Eye-tracking work at Nielsen Norman Group measures 57% of page-viewing time above the fold, 74% inside the first two screenfuls, and more than 65% of the above-fold time in the top half of the viewport (NN/g). People scroll. They just do not read what they scroll past. The offer, the proof and the action each need a version that fits in the first screen.

4. Speed measured on a phone

Google and Deloitte tracked 37 brand sites across more than 30 million sessions. A tenth of a second off mobile load time raised retail conversions by 8.4% and travel conversions by 10.1% (web.dev). One tenth of a second. A hero video, three web fonts and a tag manager pulling six scripts each cost far more than that. Before anyone argues about button colour, open the page on a mid-range Android over 4G. We wrote the Next.js half of this in how to ship green Core Web Vitals.

5. Proof next to the claim it supports

A logo wall at the bottom of the page proves nothing about the sentence at the top. Move the evidence to where the claim is: the number under the promise it backs, the quote beside the feature it describes, the security note next to the field asking for that data. Specific proof beats abundant proof. One customer saying "we cut onboarding from three weeks to four days" outperforms twelve saying "great team".

6. A form that asks for what the next step needs

The rule that three fields is optimal comes from a single blog post from the early 2010s and has circulated unchanged ever since. CXL tested it and found the opposite in a live case: cutting a nine-field form to six lowered conversions, because the removed fields were the ones prospects engaged with most, while keeping nine fields and rewriting the labels lifted conversions by 19.2% (CXL). Field count is a proxy for perceived cost, and only the second one moves. Ask for what the next conversation genuinely needs, say why for any field that feels intrusive, and validate as people type: Baymard finds 31% of sites run no inline validation at all and 4% implement it wrong (Baymard Institute).

7. Design the phone screen first

On a phone the first screen is about 640 CSS pixels tall once browser chrome is gone. That holds a headline, one supporting line and a button. Designing at 1440 and letting the grid collapse pushes the proof to 2,400px and the button to 3,100px. Start from the phone, then let the desktop version spend the extra room on evidence.

8. The objection answered before the button

Under every form sits a question nobody writes down: what happens after I click. Price range, who calls, how soon, whether this is a sales call, what happens to the data. One line under the button answering those removes more hesitation than another testimonial. The version of this that costs the most deals is in our pricing page teardown.

The 4 things that kill it

1. A hero about the company. "We are a team of engineers and designers building the future of X" says nothing to the visitor about the visitor's problem. The hero is for what they get, in their words.
2. Competing actions. Navigation bar, chat widget, cookie banner, video modal, two calls to action with equal weight. Each is a fork in a road that should be straight.
3. Friction hidden in the form. Validation only on submit, errors that wipe what was typed, required fields nobody in sales reads, a phone number demanded before anything of value has been given.
4. One page for every source. The same page serving paid search, a newsletter, a conference QR code and an AI citation converts for none of the four. Traffic from AI answers in particular lands further along the decision, and repeating the basics wastes that intent (we covered that traffic pattern here).

The order to fix them

The eight factors do not weigh the same, and teams usually start at the tail. This is the sequence we use on a page that is already live.

1. Split conversion by source. Fix the worst source, or stop buying it.
2. Rewrite the headline so it repeats the promise made by the source that sends the most traffic.
3. Remove every action that is not the primary one.
4. Measure LCP and INP on a real phone, then cut the heaviest payload until both are green.
5. Move the strongest proof next to the strongest claim.
6. Review the form field by field. For each field, name the person who will read it. Delete the ones with no name.
7. Answer "what happens next" in one line under the button.
8. Only now, change the visual design.

A page that gets the first three right and looks unremarkable beats a beautiful page that gets them wrong. That is the uncomfortable part of conversion work: the variable that is most enjoyable to tune is the one that moves the number last.

Sources

• Unbounce, Conversion Benchmark Report: SaaS conversion rates
• Nielsen Norman Group, Scrolling and Attention
• web.dev, Milliseconds Make Millions (Google and Deloitte)
• CXL, Should You Really Reduce Form Fields?
• Baymard Institute, Usability Testing of Inline Form Validation
• CXL, Give Your Advertising ROI a Serious Boost by Maintaining Scent