The eye finds your button last

You have heard the goldfish statistic: human attention span is down to eight seconds, shorter than a goldfish’s. It’s a great line for a conference slide, and it’s fiction. The BBC traced the claim back through its citations and found no study at the end of the trail, and nobody has ever timed the goldfish either.

The real research is more uncomfortable. Gloria Mark at UC Irvine has measured screen attention for two decades: in 2004 people stayed on one screen for about two and a half minutes before switching; by the time she wrote it up in her book, the average was 47 seconds. Attention isn’t broken, it’s expensive. The same people who leave your page in seconds will happily sit through a three-hour film. They spend attention on things that have earned it, and a first visit to your product has earned nothing yet.

Two more numbers frame how little runway you get. People form a judgement of a page’s design in about 50 milliseconds, before reading a word. And Chartbeat’s data across two billion visits showed a majority of visits bouncing inside 15 seconds. So the useful question isn’t “how do we say everything”. It’s: what does the eye actually do with the seconds we get? Below is the answer, drawn as one continuous take. Watch where the dot goes, and watch how long the acid button waits.

That wandering dot isn’t a parody. It’s the pattern eye-tracking labs have been replicating for twenty years. The first fixation lands near the centre, because that is where the eye already is. Then the logo, because people check where they are before they check what you want. Then reading begins, and it isn’t really reading: it’s the F-pattern, two horizontal sweeps and a lazy skim down the left edge, a finding Nielsen Norman Group first published in 2006 and keeps confirming with fresh studies.

Pictures cheat. A photograph, especially one with a face in it, pulls fixations away from everything around it. In my ads years the heatmap never matched the mockup: we would build a page around a headline and watch the photo outdraw it in test after test. Meanwhile anything that merely looks like an ad gets skipped wholesale, real content included. NN/g calls it banner blindness, and it has survived every redesign trend since the nineties. Dress your best feature in a promo costume and you’ve hidden it.

And the button? The button is usually fine: visible, well labelled, sitting in the hero where the templates say it should. It’s still confirmed last, because a call to action only gets interesting once the page has answered “what is this” and “is it for me”. The eye doesn’t hunt for buttons. It hunts for reasons, and it collects them from the whole screen first.

Phones change the geometry but not the physics. On a small screen the F-pattern collapses into a straight vertical skim, thumbs scroll faster than eyes fixate, and the attention budget shrinks further. What survives the move is the ranking of what wins a glance: faces first, then contrast, then whatever moves. Desktop is where the wandering gets expensive, because there’s so much more surface to wander across, which is exactly why the animation above is drawn on a big screen.

Every screen has a contrast budget. One element can be loud. The moment two things shout, neither is heard, and the eye quietly files the whole page under noise. My rule from performance creative: decide the single thing the glance must carry, give it the accent, and demote everything else without mercy. Most pages I audit fail this test before any other.

Isolation beats size. Fitts’s law says bigger and closer targets are faster to hit, and it’s true, but on a crowded page the fastest button is the one sitting alone in clean space. Whitespace isn’t emptiness; it’s targeting. A smaller button with room around it beats a bigger one wedged between competing blocks.

Then put the payoff where fixations land anyway. The F-pattern is a gravity map: front-load your headline with the words that matter, start list items with the noun that carries the meaning, and let the call to action sit at the end of the natural reading line instead of in a decorative corner. A KYC step, a pricing card, an empty state: each is a small landing page with one job, which is the argument of what most fintech onboarding gets wrong in different clothes.

The cheapest test remains the blur test: step back, squint, and see what survives. If the blurred page doesn’t show one obvious next step, the sharp one doesn’t either; sharpness adds detail, never hierarchy. That’s the two-second discipline from every screen is an ad, applied at desktop scale.

I’ll admit where contrast budgets actually die, though: in review meetings, one reasonable request at a time. Nobody sets out to build a page with four primary buttons. Sales needs the demo link, growth needs the trial, someone senior loves the video, and each ask is sensible on its own. The page ends up shouting in four directions, and the eye, which never attended the meeting, files the whole thing under noise. Half of guarding a design is doing the arithmetic out loud: we can afford one loud thing. Which one is it?