Answer in brief
A landing page works when a first-time visitor can name the offer, the audience and the next step after a few seconds. Check it with a five-second test: show the first screen briefly, hide it and ask open questions. Then make sure the main content appears quickly (web.dev calls a Largest Contentful Paint of 2.5 seconds or less good) and that text contrast, keyboard access and button size meet the basics of WCAG 2.2.
A stranger decides in seconds whether the page earns a minute
A visitor reaches a landing page with a question, from an advert, a search result or a link in a message. Nobody owes the page patience. In a 2011 article for the Nielsen Norman Group, Jakob Nielsen summed up a Microsoft Research analysis of page visits: users often leave web pages in 10–20 seconds, but pages with a clear value proposition can hold attention much longer.
The detail matters more than the average. According to that article, the first 10 seconds are critical for the decision to stay or leave, and the chance of leaving remains high for the next 20. Nielsen’s conclusion: the value proposition must be communicated clearly within 10 seconds.
So a useful landing page critique does not start with colour, animation or taste. It starts with three things a newcomer should be able to state almost at once: what is offered, for whom, and which step comes next. The figures describe browsing in general, not a stopwatch rule, but they set the order of work: the answer first, then whatever lets it be seen.
Running a five-second test the way researchers describe it
The Nielsen Norman Group’s guide to testing visual design calls the five-second test a first-impression test: you show a design for five seconds to capture a gut reaction. That is too short for reading copy, the guide says, but enough to form an impression of the visual style. The same brief exposure suits a harder question: whether the offer has been understood.
The setup takes minutes. Take a screenshot of the first screen as it appears on a phone. Find a handful of people who resemble your customers and have never seen the page, which rules out colleagues. The guide adds one instruction: do not warn participants that the image will vanish after five seconds, because the warning primes them to memorise details.
Show the screenshot, hide it and ask open questions in a fixed order. What does this company offer? Who is it for? What would you do next? Write the answers down word for word. The guide advises against pairing so short an exposure with a long list of adjectives: by the end of the list people no longer remember much of what they saw.
Scoring the answers without flattering your own page
Before the first session, write the answer you hope to hear in one sentence, for example “accounting software for freelancers, free to try”. An answer passes when the person names the offer and the audience in words close to yours. It fails when they name the wrong category, describe decoration (“something with a blue gradient”) or retreat into “some kind of solutions”.
The pattern of failures tells you what to change. People who recall the picture but not the offer saw a visual that outweighs the message. People who repeat the headline but cannot say what to do next did not register the button as the main action. Answers that differ widely between people usually mean the first screen says several things at once.
A five-second test measures the clarity of a first impression, not persuasion: it cannot tell you whether the price is right or the form works. The same guide describes a companion, the first-click test, in which a participant receives a task and is stopped after the first click. It shows whether the next step can be found.
What the first screen must carry and what can wait
Nobody scrolls in five seconds, so the test is really a test of the first screen. That screen needs four things: a headline that names the offer in the customer’s words; a line saying who it is for or what result to expect; one primary action whose label states what happens after the click; and one piece of checkable evidence, such as a product screenshot or a named client.
Almost everything else can wait below: the company history, the full feature list, secondary offers. Lower sections serve visitors who already know what the page is about and want reasons. But the first screen must not look like the whole page. A hero that fills the display edge to edge, with no hint of what follows, can read as the end of the story.
Judge the phone version first, because that screen is the smallest. Once a consent banner, a sticky header and a chat bubble have taken their share, the headline may sit below the visible area and the button may be gone. Take the test screenshot on a real phone, with the banner as a new visitor meets it, not from the tidy mock-up in the design file.
Largest Contentful Paint: when the answer actually appears
The visitor’s seconds start at the tap, not when the page has finished loading. Google’s web.dev documentation measures this with Largest Contentful Paint, or LCP: the render time of the largest image, text block or video visible in the viewport, counted from the moment the user first navigated to the page. It marks the point at which the main content has probably loaded.
Sites should strive, web.dev says, for an LCP of 2.5 seconds or less, measured at the 75th percentile of page loads and separately for mobile and desktop. In other words, three visits in four should show the main content within that time. If a hero image needs four seconds, to take an invented example, most of the first five seconds pass before the offer appears.
The same page lists where to read the number: PageSpeed Insights and Search Console for data from real visits, Lighthouse and Chrome DevTools for a laboratory measurement. On a landing page, look first at the hero image and the display fonts: an image counts only once it has loaded, the documentation notes, and text in a web font is not treated as rendered during the font block period.
Accessibility basics that let the answer be read at all
A message that cannot be read has not been delivered. The Web Content Accessibility Guidelines (WCAG) 2.2, a W3C Recommendation, contain criteria that bear directly on the first screen. The best known is contrast: text needs a ratio of at least 4.5:1 against its background, large text at least 3:1 (criterion 1.4.3, level AA). Pale grey on white and words laid over a photograph are the usual failures.
Next come images and the keyboard. Non-text content needs a text alternative that serves the equivalent purpose (1.1.1, level A), so an offer that exists only inside a banner image is silent for anyone using a screen reader. All functionality must be operable through a keyboard (2.1.1), with a visible focus indicator (2.4.7). To check, press Tab until you reach the main button, then press Enter.
Version 2.2 added a minimum target size of 24 by 24 CSS pixels for pointer inputs, with exceptions such as links inside a sentence (2.5.8, level AA), and fields that require input need labels or instructions (3.3.2). These quick checks are not a conformance audit. Whether a business is legally obliged to meet WCAG depends on national law, a question for a qualified adviser.
Design tricks that stand between the visitor and the answer
The rotating banner is the classic case. A carousel puts several messages in the place of one, and in a five-second test some participants may describe the second slide while your offer sits on the first. WCAG has a rule for it: moving content that starts automatically, lasts more than five seconds and runs alongside other content needs a way to pause, stop or hide it, unless the motion is essential (2.2.2, level A).
Entrance effects cost time more quietly. A headline that fades in after a two-second animation has wasted two of the visitor’s five seconds. A large hero video can be the element LCP measures; web.dev notes that for video the poster image or the first frame counts, whichever comes earlier. And a clever headline that could belong to any company (“Solutions for a better tomorrow”) answers nothing.
One test separates decoration from obstruction: does the effect help a stranger answer one of the three questions sooner? If not, move it below the first screen or remove it. This is no argument against craft. The Nielsen Norman Group guide observes that a strong first impression can build trust and interest, and a calm, well-made first screen does exactly that.
A repeatable routine: the order of checks and of fixes
Put the checks in one sequence and a landing page critique fits into an afternoon. Write the sentence you want visitors to repeat. Run the five-second test on a phone screenshot, then the first-click test. Look up LCP in field data. Finish with contrast, text alternatives, keyboard and target size.
Fix problems in the order in which they block the answer. Start with what cannot be read, which means low contrast and text locked inside images. Move on to what is unclear: the headline, the supporting line and the button label. Then deal with what is slow, beginning with the element LCP reports as the largest. Motion and decoration come last.
Retest with new people, because anyone who has seen the page is no longer a first-time visitor. Keep a dated record of each round: the screenshot, the questions and the answers as spoken. The critique then stops being a contest of opinions and becomes a comparison of versions. When strangers repeat your sentence back to you, the page is ready for an A/B experiment, which measures persuasion, not clarity.
Practical checklist
- Write one sentence stating the offer, the audience and the action you want, before showing the page to anyone.
- Show a phone screenshot of the first screen for five seconds to first-time viewers, then record their answers word for word.
- Look up the page’s Largest Contentful Paint in field data and compare it with the 2.5 seconds that web.dev calls good.
- Check headline and button text against the 4.5:1 contrast ratio in WCAG 2.2 and reach the main action using only the keyboard.
- Move below the first screen, or remove, every moving element that does not make the offer clearer to a newcomer.
Questions and answers
What is a five-second test for a landing page?
It is a quick first-impression check. You show a screenshot of the page to someone unfamiliar with it, take it away after five seconds and ask what the company sells, whom it serves and what they would click. The Nielsen Norman Group describes the method as a way to capture a gut reaction: the time is too short for reading copy, so only what is immediately clear survives.
How many people should take part in a five-second test?
The sources used here describe the method without prescribing a number. As a working rule, a handful of people from your intended audience is enough to reveal a pattern, because the test is qualitative: you are looking for repeated misunderstandings, not percentages. If the same wrong answer comes from several people, you have a finding. Use new participants each time the page changes.
What belongs above the fold on a landing page?
Only what a first-time visitor needs in order to decide whether to stay: a headline naming the offer, a sentence about the audience or the expected result, one clearly labelled primary action and one piece of checkable evidence. Features, company background and secondary offers belong further down. Verify the selection on a phone, where banners and sticky bars leave far less room than a design mock-up suggests.
What is a good Largest Contentful Paint score?
Google’s web.dev documentation gives 2.5 seconds or less as the mark of a good user experience. The yardstick is the 75th percentile, so at least three page loads in four should meet it, with phones and desktops taken apart. The figure refers to real visits, so read it from field tools such as PageSpeed Insights or Search Console. Laboratory tools such as Lighthouse are useful for finding the cause of a slow result.
Does a five-second test replace A/B testing?
No, the two answer different questions. A five-second test shows whether strangers understand the page at a glance, and it can be run on a draft before any traffic is bought. An A/B experiment compares how two live versions perform with real visitors, which says something about persuasion. Clarity comes first: a version nobody understands is not worth sending traffic to.