Five Second Test

The team that built a landing page can't judge how it reads to someone seeing it new. By the twelfth revision of the headline they've looked at it hundreds of times. Ask them whether it's obvious what the service does and the question doesn't even land, because to them it is obvious.

Asking the person at the next desk didn't help much either. At Pagedeck, Ravi would put a new landing page mockup on screen and say "you can tell what this does, right?" — and everyone nodded. Every person in that room already knew the product, so the nod came from memory, not from the screen. Even bringing in someone who had never seen it produced a considered answer after twenty seconds of reading every sentence. Real visitors don't read that way.

The five second test solves this by taking the time away. It cuts exposure down to something closer to what a real visitor gives a page, then hides the screen before asking, so there's no re-reading. The answers change once you do that, because what a page yields under careful reading and what survives a glance are different things. Within UX research, the impression branch is the one that deals with those first few seconds rather than with judgments formed after long use.

Preparation, session, and tally. A full round takes about two days.

1. Write down what you want to know, in one sentence. "Evaluate the new landing page" isn't a question. "Does a first-time visitor recognize what this page sells?" is.
2. Pick one screen. Crop it to the above the fold area — what's visible before any scrolling — and flatten it to a static image. Leave links live and participants wander, which means no two people saw the same thing.
3. Recruit around twenty people who don't know the product. Answers scatter and the finding is how many of them said the same thing, so five people won't settle anything.
4. Show it for five seconds, then hide it. Online tools handle the timing and the cover automatically; running it yourself over a screen share works too.
5. Ask the questions in order, right after the screen goes away. The first one has to carry no hint at all — "what do you remember?" Questions about what the company sells and who it's for come after that. Reverse the order and the question hands over the answer. "It seemed like a collaboration tool, didn't it?" is the most common way this goes wrong.
6. Group similar answers and count them. Fourteen of twenty saying "a document tool a team shares" means it landed; five means it didn't. What the other six said points at what to fix.

Set the bar before you run it. Decide how many counts as a pass after seeing the results and any number reads as a pass.

It fits any screen that has to convey something on first sight: landing pages, the top of a pricing page, the first screenshot in an app store, ad creative. It's also where you find out whether the value proposition — the compressed sentence for why a customer should pick this product — is actually being read.

It cannot tell you whether people can complete a task. To see where users get stuck on an invite screen you need a usability test, which puts a real task in front of someone; to see where the first click goes, a first click test. It also doesn't suit feature screens used daily by existing customers, for whom first impressions are long past.

There are three limits. First, memorable isn't the same as important. An illustration covering half the screen is remembered well and has nothing to do with signups, so what people recalled and what needed to land are two separate lists.

Second, it reports that something didn't land without explaining why. Five correct answers instead of fourteen doesn't say whether the wording or the type size caused it. The result is usually the start of the next study rather than a conclusion.

Third, the number five has no great authority behind it, and real visitors sometimes leave faster. What matters is holding the value steady across the team. Measure at eight seconds last time and five this time and the two results can't sit side by side.

"Let's run a five second test on this headline. We can't judge it ourselves."

The people who made it already know what it says, so their first impression doesn't exist.

"Fourteen out of twenty said document tool. I think we can leave it."

The pre-agreed bar was cleared, so this screen is done.

"All they remember is the logo and the photo of people."

The most visible element is crowding out the thing that needed to land.

Pagedeck used this while reworking its pricing page. Ravi ran the existing screen first: seventeen of twenty recalled the three plan names, and four could say how the plans differed. The table underneath held twelve rows, and five seconds only bought the headings.

Noa left the table alone and added a one-line summary under each plan. On the rerun, twelve people could name a difference. Ravi was careful not to claim this would raise purchases, though. Landing and buying are separate questions, and he checked the second one after the change actually shipped.

The most common mistake is treating it as a way to pick between two designs. The shape is similar enough — show something briefly, collect reactions — that it looks like a comparison tool. But the only answer it produces is whether the message landed, not which option is better. Run two mockups separately and start comparing scores and you end up choosing a design on a number that flips when the participants change or the question order shifts slightly. If picking is the goal, put them side by side from the start.

The other is believing five participants is enough. The rule that five people surface most major problems in usability testing is well known, so it gets carried over. The two methods count different things. Usability testing looks for places people get stuck, and one person stuck is enough to expose one. This measures what share of people got the message. At five participants, one person moves the result by twenty percentage points, so the same screen tested twice gives different answers.