Does a stranger understand your homepage in five seconds?

1. Blog
2. Conversion & Pricing

By Saroj Jha · November 2, 2026 · 13 min read

Most startup homepages fail a simple check: a stranger who sees the top of the page for five seconds cannot say what the product is, who it is for, or why it beats what they do today. You can find out in an afternoon. Show a screenshot of your hero to five to ten people who match your buyer but do not know you, hide it after five seconds, ask those three questions, and score each answer 2, 1 or 0. Rewrite whatever most people missed, in the order what, who, why, then test again with new strangers. A five-second test tells you whether the main message lands; it does not tell you whether the page converts.

What follows applies those findings to one fictional company’s homepage.

Running case: Example Co. Fictional

Example Co. is a fictional seed-stage startup whose AI tool drafts answers to security questionnaires for B2B SaaS companies, using their past answers and policies. Its buyers are sales engineers and security leads who today copy answers from old spreadsheets. Its homepage hero reads: “The AI-native trust platform for modern teams,” with a “Book a demo” button. Demo requests from the homepage are low, and the founder wants to know whether visitors even understand what the product does. Every Example Co. number here is fictional and its arithmetic is shown.

What does the top of a homepage have to answer?

In my view the hero, meaning the headline, the line under it and the first visual, has three jobs. It must say what the product is, in words a buyer would use. It must say who it is for, so the right visitor sees the page is for them and the wrong one leaves. And it must say why it beats the alternative the visitor uses today, which is often a spreadsheet, an agency or doing nothing.

Everything else on the page can wait. Logos, feature lists and pricing all assume the visitor has already understood the first three answers. A hero that is clever, abstract or all category language (“platform”, “AI-native”, “solutions”) usually fails the first question, and no amount of proof lower down rescues a visitor who has already left.

The what and who come from your positioning; I covered how to write them in Value Proposition and Messaging Framework. This post is about checking whether the result lands with someone who has never heard of you.

Example Co., step 1 Fictional

“The AI-native trust platform for modern teams” names no task (what), no buyer (who) and no alternative (why). Against the three jobs, the hero answers none of them.

Do this

Write your current hero on one line, then write next to it the one-sentence answer you want a stranger to give to each of the three questions. If you cannot write those three answers, fix the positioning before the page.

Why do the first few seconds decide so much?

Visitors judge a page fast and leave fast. Researchers at Microsoft Research analyzed two weeks of browsing logs from opted-in users of a US English browser plug-in, covering 205,873 pages with at least 10,000 visits each, and found that on 98.5% of pages the time people stayed followed a pattern where the chance of leaving is highest early and falls the longer someone stays^[1]. During that early screening, in the authors’ words, the probability of abandoning the page is high^[1]. They also found that the typical (fitted) dwell time was 70 seconds or less on 80% of those pages^[1].

Jakob Nielsen of Nielsen Norman Group drew the practical conclusion in 2011: users often leave web pages in 10 to 20 seconds, and to gain several minutes of attention you must clearly communicate your value proposition within 10 seconds^[4]. The study measured all kinds of web pages, not B2B homepages, so I treat it as a reason to care about the first glance rather than as a benchmark for your page.

The five-second test checks exactly that first glance. Guillaume Gronier’s 2016 peer-reviewed study cites Christine Perfetti (2007) and Alabi (2007) as describing the protocol, and notes that before his work the test had rested mainly on practitioner blogs rather than research^[2].

What does the research say a five-second test can and cannot tell you?

Gronier’s study, published in the Journal of Usability Studies, randomly assigned 146 people to see a weather website’s homepage either for exactly five seconds (68 people) or for as long as they liked (78 people)^[2]. After five seconds, 97.1% correctly named the site’s main purpose, current weather, but only 2.9% named weather warnings and none named five-day forecasts^[2]. The unlimited-time group did no better on purposes overall: mean scores were 1.00 out of 3 for the five-second group and 1.05 for the unlimited-time group, a difference that was not statistically significant^[2].

Two lessons follow for a homepage. First, five seconds is enough for people to pick up the one main message, so the hero should carry one message and not three. Second, five seconds is not enough to judge everything: Gronier concludes that the test is not enough time to let users identify all the objectives of a website, and the unlimited-time group rated information quality and perceived ease of use higher, although the difference in the overall usability score was not significant^[2]. So I use the test only for the hero’s three answers, never as a verdict on the whole page.

Example Co., step 2 Fictional

The founder decides the test will check three things only: what the product is, who it is for and why it beats copying from old spreadsheets. Features, pricing and integrations stay out of the test, because five seconds will not show whether visitors understood them.

Do this

Write down, before you test, the single main message you want a stranger to repeat back. If you have two or three, choose one for the hero and move the others down the page.

How do you run a five-second test on five to ten strangers?

Recruit people who match your buyer but have never seen your product: not your team, not investors, not friends who know what you do. For a B2B product that might be people in the same role at companies outside your pipeline, reached through your network, a community or a paid research panel. Five to ten is enough for a first read; Nielsen’s model of usability testing found that five comparable users uncover about 85% of usability problems, with each user finding about 31% on average across the projects studied^[3].

That model is about usability problems, not message comprehension, so I borrow its logic rather than its number: a handful of comparable people shows the big failures, and adding more mostly repeats them. Nielsen also notes the model holds only for comparable users, and recommends three or four people per group when you test two distinct groups^[3]. If your homepage serves two buyers, test each separately.

The method takes about ten minutes a person:

• Take a screenshot of the hero at real size, on desktop and on mobile, cropped where the visitor’s screen ends.
• On a video call or in an unmoderated testing tool, show it for five seconds, timed, then hide it.
• Ask three open questions in this order: “What does this company sell?”, “Who is it for?”, “Why would someone choose it over what they use today?”
• Write down the exact words of each answer; do not prompt or correct.
• Ask last: “What would you click first, and why?”

Example Co., step 3 Fictional

The founder recruits eight sales engineers and security analysts from a community forum, none of whom know Example Co., and runs the test over video calls with the desktop hero, five seconds each.

Do this

Book five to ten 10-minute calls this week with people who match your buyer and do not know you, and prepare the screenshot and the three questions before the first call.

How do you score the answers?

Score each answer against the one-sentence answer you wrote in step 1, not against how kind the person was. My scale is 2 for correct in substance, 1 for partly right or vague, and 0 for wrong or “I don’t know.” Each person can score up to 6, and each question up to twice the number of people.

Then count how many people got each question fully right. My rule of thumb, not a benchmark, is that the hero passes when nearly everyone gets “what” right (at least 4 of 5, or 7 of 8), most get “who” right (at least 6 of 8) and at least half get “why” right. Gronier’s 97.1% for a main purpose^[2] shows what a clear main message can reach, though his page was a weather site, not a B2B product.

Read the wrong answers as carefully as the scores. When several strangers say the same wrong thing (“some kind of compliance auditor”), that phrase tells you which word in your hero is misleading them.

Example Co., step 4 Fictional

Eight people, three questions, scores of 0 to 2. “What”: 2 of 8 fully right, 6 of 16 points (38%). “Who”: 1 of 8, 4 of 16 points (25%). “Why”: 0 of 8, 1 of 16 points (6%). Total 11 of 48 points, 23%. Three people guessed “a compliance auditing firm,” pointing at the word “trust.”

Do this

Put the answers in a sheet with one row per person and one column per question, score 0, 1 or 2, and list the repeated wrong phrases under the scores.

How do you rewrite the hero, and how do you know it worked?

Fix the questions in order: what, then who, then why. A visitor who does not know what you sell will not care who it is for. Make the headline name the task in the buyer’s words, put the buyer and the mechanism in the line under it, and state the alternative you beat in plain terms, with a number only if you can prove it.

Take the words from buyers, not from your pitch deck. Sales call notes, support emails and the strangers’ own answers are the best source; the phrases people used when they guessed right are often better than your draft. The Positioning Statement Generator helps when the gap is in the positioning rather than the wording.

Then test again with new strangers, never the same ones, because people who saw version one already know the answer. Once the hero passes, the question changes from “do they understand?” to “do they act?”, which needs real traffic and an A/B test; the A/B Testing guide covers when you have enough visitors for that.

Example Co., step 5 Fictional

New hero: “Answer security questionnaires in hours, not weeks.” Line under it: “Example Co. drafts answers from your past responses and policies, for B2B SaaS sales and security teams who are done copying from old spreadsheets.” Retest with eight new strangers: “what” 7 of 8 fully right (15 of 16 points, 94%), “who” 6 of 8 (13 of 16, 81%), “why” 5 of 8 (12 of 16, 75%); total 40 of 48 points, 83%, against 23% before. It passes all three of the founder’s rules. The “hours, not weeks” claim stays only because Example Co. has customer data to back it.

Do this

Rewrite only the question that failed first, retest with new people, and keep a log of each version and its scores.

When is a five-second test the wrong tool?

It is the wrong tool for anything beyond the main message. It cannot judge navigation, pricing, or whether a visitor will find the integration page, which need longer usability sessions; Gronier’s results show five seconds misses secondary purposes^[2]. It also cannot predict conversion: a hero everyone understands can still describe a product nobody needs.

It is also the wrong tool for paid landing pages judged on their own. A visitor from an ad arrives with the ad’s promise in mind, so that page has to match the ad as well as be clear, which I covered in Why Your Landing Page Converts at 1%. And a five-second test on people outside your market tells you about general clarity, not about your buyer; recruit the right strangers or treat the result as a rough check.

In my view the right moment for a five-second test is before every hero rewrite and after every repositioning. It is cheap, it is fast, and it catches the most expensive homepage mistake: a message only the founders understand.

What changes by stage?

• Pre-seed — Test the hero before you build the rest of the site, with five people from your target role. The positioning is still moving, so expect two or three rounds.
• Seed — Retest whenever the positioning or main segment changes, and test mobile separately. Once the hero passes, move to A/B tests on the homepage if traffic allows.
• Series A–B — Test each audience’s version of the hero with its own strangers, three or four people per group when you serve several buyers, and fold the test into every site redesign.

Free tool

Score what strangers remember in five seconds

The Five-Second Message Test runs this test for you: it shows your screenshot to testers for five seconds, asks what they remember, scores each answer against the message you intended, and gives the share who got each question right with a confidence interval, flagging results from fewer than five people as a first look.

Before the hero, settle the positioning with Value Proposition and Messaging Framework and the Positioning & Messaging Workbook. After it passes, the rest of the page is covered in the Website & Conversion Optimization Playbook.

If you want me to look at your own funnel, positioning and channels, my Growth Audit ends with three prioritized fixes and a 30-day action plan.