How To Read A Heatmap And A Session Recording

A dashboard tells you that sixty percent of people leave your service page. It will never tell you that they leave because the price section looks like a navigation menu and they clicked it expecting a link.

That you find by watching somebody do it.

What each one is good for

Scroll maps show how far down the page people get. The single most useful one, because it tells you what is effectively not on the page at all. If eighty percent never reach your proof section, your proof does not exist.

Click maps show where people click. Useful mostly for what they reveal about expectations: people clicking things that are not links, people clicking the same button twice because nothing visibly happened.

Move maps show cursor movement. The least reliable of the three. Treat them as a hint, not evidence.

Session recordings show one real person using the site. The most valuable and the one nobody watches, because it takes time.

Watch ten sessions, not a hundred

You do not need volume here. You need attention.

Pick ten recordings from people who did the thing you care about, and ten from people who left. Watch them at normal speed, not sped up, with a notepad.

Patterns show up in the first five. By ten you will have three or four specific problems you had no idea existed. More than that is diminishing returns until you have fixed those.

What to look for

Rapid scrolling up and down. They are looking for something and cannot find it. Usually price, usually service area, usually a phone number.

A long pause partway down. Either they are reading something closely, which is good, or they are stuck, which is not. Watch what happens next: reading leads to continuing, stuck leads to leaving.

Clicks on things that are not clickable. An image they expected to enlarge, a heading they expected to expand, a logo they expected to link. Each one is a small broken promise.

Rage clicking. Three or four clicks in the same spot in a second. Something is not responding and they think it is their fault. These are always worth fixing and usually quick.

Form abandonment, field by field. Where in the form did they stop. Phone number and company size are the two usual culprits, and that is the argument in why your enquiry form is too long.

The back button after a click. They clicked through, found the wrong thing, and returned. Your link text promised something the page did not deliver.

On mobile, how much of the first screen is taken by a cookie banner, a chat bubble and a sticky header. Watch a mobile session and you will see how little of the actual page anybody gets.

Do mobile separately

Mobile and desktop behave differently enough that averaging them hides everything.

Most traffic is mobile and most sites are designed on a laptop. Watching ten mobile recordings is usually the single highest return hour available, because the problems are large and obvious the moment you see them.

Turn it into a change, then measure it

The point is not insight. It is a change you can test.

So every finding becomes one specific edit: move the proof above the fold, cut three form fields, make that heading a real link, put the phone number in the header. One change at a time where you can, so you know which one did it.

And then you check it in the numbers, which only works if the tracking is sound, as in how to set up conversion tracking properly.

Where it fits

This is the step between knowing a page underperforms and knowing what to do about it. It sits directly between how to audit your own website and how to fix a page that gets traffic but no leads.

And it answers the question a lot of founders ask in the wrong order. Before you rebuild the site, watch twenty people use the one you have. Half the time the rebuild turns out to be four edits, and the other half you at least know what the new one has to fix, which is where what makes someone trust a website starts.