First-click testing shows people a screen, gives them a task such as "Where would you go to find running shoes?", and records where they click first.
It matters because the first click predicts the rest of the journey: in Bob Bailey and Cari Wolfson's studies, people whose first click was correct finished the task 87% of the time, against 46% when it was wrong.
Five-second tests and preference tests are its cheap cousins, checking first impressions and stated taste before you spend traffic on an A/B test.
All three are pre-test research methods. They tell you where a design is confusing.
They do not tell you what it will do to revenue, and our database has several tests where the click data and the sales data went in opposite directions.
First-click testing: how it works and why the first click matters
You upload a screenshot, wireframe, prototype or live page, write a task, and mark which areas count as a correct click.
Each participant gets the task and clicks once.
The tool reports where they clicked, as a heatmap, plus how long they took and what share hit the right area.
Lyssna's first-click feature reports all three and suggests 15 to 20 participants to find major navigation problems.
The research behind the method comes from Bailey and Wolfson's work across a dozen scenario-based studies between 2007 and 2009. Jeff Sauro at MeasuringU summarized those numbers and then checked them against MeasuringU's own data. The gap depended on context:
| Data set | Success after a correct first click | Success after a wrong first click |
|---|---|---|
| Bailey and Wolfson, 12 studies | 87% | 46% |
| MeasuringU live websites (5 sites, 750 users, 21 tasks) | 78% | 70% |
| MeasuringU tree tests (8 studies, 1,000+ users, 137 tasks) | 80% | 14% |
The lesson is that the first click matters most when navigation is the only route to the answer, as in a tree test or card-sort follow-up.
On real sites with search and several paths, a wrong first click is easier to recover from.
Some vendor pages say users are "87% more likely to succeed" after a correct first click.
That misreads the study: 87% is the success rate, not the size of the increase.
First-click tests suit menus, category pages, homepages and any screen where people have to pick a path.
They are the fastest way to check a new navigation menu before you build it.
Preference testing: what it can and can't tell you
A preference test shows people two or more designs side by side and asks which they prefer and why.
Lyssna lets you compare up to six designs, works out statistical significance when there are two, and suggests 30 to 50 participants for quantitative results.
The weakness is the one Jakob Nielsen pointed out in 2001: pay attention to what users do, not what they say.
In a later update to that article, Nielsen reports a correlation of r = .53 between how much people liked websites and how well they performed on them.
Squaring that gives about 0.28, so stated liking predicts only about a quarter of how well a design actually works.
Small samples also make a strong-looking result unreliable.
Here is a worked example using an exact two-sided binomial test, where the null hypothesis is that people choose each design 50% of the time:
| Participants | Chose design A | Share | p-value | Verdict at 0.05 |
|---|---|---|---|---|
| 30 | 20 | 67% | 0.099 | Not significant |
| 30 | 21 | 70% | 0.043 | Significant |
| 50 | 30 | 60% | 0.203 | Not significant |
| 50 | 32 | 64% | 0.065 | Not significant |
The p-value is 2 × P(X ≥ k) for X ~ Binomial(n, 0.5).
A two-to-one split among 30 people still isn't clear evidence.
Even a significant result only tells you which design people say they like, which is not the same as which one they buy from.
Use preference tests for brand calls where taste really is the goal: a logo, packaging, an ad visual, a choice between two photo styles.
They are the wrong tool for deciding between two checkout or product page layouts. Run an A/B test for those.
The 5-second test: checking the first impression
A 5-second test shows a page for five seconds, hides it, then asks what people remember.
Christine Perfetti described it at User Interface Engineering in "5-Second Tests: Measuring Your Site's Content Pages".
The time limit is the point. As Perfetti put it, when people get longer than five seconds, "they start looking at the page more like a designer, noticing details they would normally miss or misinterpret."
Perfetti's example was the Red Cross donation page.
Every participant knew the page was about donating, but none noticed they could give stock, clothes or airline miles.
A redesign listed the options at the top, and in a repeat test participants recalled the full set after five seconds.
The idea that first impressions form fast has research behind it.
Gitte Lindgaard and colleagues found that ratings of a homepage's visual appeal after a 50-millisecond view closely matched ratings after 500 milliseconds (Behaviour & Information Technology, 2006).
That study measured how appealing a page looked, not whether people understood it or bought from it.
Don't cite it as proof that a page has 50 milliseconds to make a sale.
How to run a five-second test that gives usable answers
- Test one question. Pick the job the page has to do in five seconds: say what you sell, show the price, or show what to click next. Perfetti's version starts with a focused task, then shows the page.
- Use the real first screen. Test the mobile above-the-fold view if most traffic is mobile, not a full-length desktop mockup.
- Ask open questions first. "What does this company sell?" and "What would you do next?" come before any leading question. Then ask for one word that describes the brand.
- Score against a written answer. Decide beforehand what a correct answer to "What does this sell?" looks like, then count the matches.
- Pick a sample size for the goal. Lyssna suggests 5 to 10 people for qualitative findings and 30 to 50 for quantitative ones.
- Recruit people who look like your buyers. A generic panel will misread jargon your customers understand. For B2B pages, a targeted panel such as the ones on our message testing tools page is worth paying for.
If most people can't say what you sell, you have found a value proposition problem worth fixing before any A/B test.
If they can, the five-second test has done its job and can't tell you anything more.
What our A/B tests show about clicks, preferences and revenue
These methods measure clicks and opinions. The tests in our database show those two things don't reliably predict sales.
Clean Commit tested a decluttered header and fuller footer on a sports equipment store.
Click-through rose 12%, but conversion rate fell 16.3% and revenue per visitor fell 16.5% across 36,149 visitors over 35 days.
The result wasn't significant (the original had a 91.0% probability of being better), but more clicks clearly didn't mean more orders.
The reverse happened on a homepage rebuilt around product clarity, also from Clean Commit.
Click-through fell 15.5% while conversion rate rose 11.0% and revenue per visitor rose 21.4%.
That one was directional too: 23,885 visitors and a 68.1% probability of beating the control.
A first-click study scoring clicks would have ranked these two designs the wrong way round.
Preference is just as risky. A single-row header with an image mega-menu looks like the kind of design people pick in a side-by-side.
Against the old multi-row header it cut conversion rate by 22.5% and orders by 22.7% across 10,957 mobile visitors, and Clean Commit gave the control a 99.5% probability of being better.
Conversion Fanatics saw a similar pattern with a luxury skincare kit page redesign: better images, a benefit headline, social proof and a visible price, and it lost by a large margin.
The wins come from using behavior data to choose what to test.
Oliver Kenyon found that most users went through the menu first and that search had the highest revenue per click, so the team put a search bar inside the mega menu.
The test added $81,050 a month. Dave Diederen used heatmaps and scroll data to move the USPs buyers engaged with above the fold, for +6.23% conversion rate and +7.28% revenue per visitor.
Kasper Dolk took objections from customer surveys and answered them in the product hero, reporting $763,679.14 extra profit a year.
The process works best end to end. Surefoot's LovelySkin redesign went through three rounds of user interviews and ten pre-launch A/B tests on its riskiest parts, and conversion rate rose 39% after launch.
That figure is a before-and-after comparison, not a controlled test.
@oliverkenyon
@DaveDiederen
@kasper_dolk
Clean CommitClean CommitClean CommitWhich method to use, and when each is the wrong choice
| Method | Answers | Typical sample | Wrong choice when |
|---|---|---|---|
| First-click test | Can people find the right path? | 15-20 to find major issues | You want to know whether a layout sells more, or the page has no navigation decision |
| Five-second test | Is the main message clear at a glance? | 5-10 qualitative, 30-50 quantitative | The page is meant to be read slowly (long-form sales pages, pricing details, forms) |
| Preference test | Which option do people say they like? | 30-50 | The choice affects conversion and you have the traffic to A/B test it |
| A/B test | Which version changes real behavior? | Set with a sample size calculator | Traffic is too low to reach significance in a few weeks |
A simple rule: use first-click and five-second tests to find problems, use full usability testing to understand them, and use A/B tests to decide.
Behavior data from heatmap analysis and answers from customer surveys are often a better source of test ideas than any panel, because they come from real buyers.
If traffic is too low for an A/B test, a first-click or five-second test is better than a guess.
Treat the result as a reason to ship carefully and watch the numbers, not as proof of a lift.
Our list of best practices that failed shows how often confident predictions lose.
For the wider toolkit, see the customer research hub or browse all tests.