Here's a failure mode I see constantly: someone stares at a Google Analytics funnel, spots a big drop-off on one step, and immediately "fixes" it — reworks the page, moves the button, changes the copy. Weeks of work. And the number barely moves, because they never actually watched a single human being use the thing. They optimised a hypothesis they'd invented from a bar chart.

Google Analytics is superb at telling you what happened and useless at telling you why. Hotjar is the opposite. The trick is using them together, in the right order, so you stop guessing.

What each tool is actually good at

Google Analytics is your quantitative layer. It's the scoreboard: sessions, sources, the funnel, where people enter and where they leave, which pages carry the most weight. It tells you where to look. What it can't tell you is what was going through someone's head when they abandoned a form at field three.

Hotjar is your qualitative layer. Heatmaps show you where people actually click, move and scroll — often nowhere near where you assumed. Session recordings let you watch real (anonymised) visitors move through the site. On-page polls let you literally ask "what stopped you today?" at the moment of hesitation. It tells you why.

Neither is enough alone. GA without Hotjar is a diagnosis with no examination. Hotjar without GA is a handful of vivid anecdotes with no sense of whether they're representative.

The workflow that actually works

The order matters. Use the quantitative tool to find the where, then the qualitative tool to understand the why.

1. Let GA point at the wound

Start in Analytics. Build the funnel for whatever matters — sign-up, checkout, contact. Find the step with the ugliest drop-off. Don't theorise about it yet. All you've established is a location: "we lose a big chunk of people between the pricing page and the form." That's a where, not a why.

2. Go and watch

Now open Hotjar and look at that exact step.

  • Heatmap first. Are people even seeing the call-to-action? Click maps regularly reveal that the button everyone assumed was obvious sits below the fold, and half your visitors never scroll to it. Move maps show where attention actually goes. More than once I've watched people click something that isn't a link — which tells you precisely what they expected to be clickable.
  • Then recordings. Filter to sessions that hit the problem page and left. Watch ten of them. You're looking for the moment of friction: the rage-click on a dead element, the long hover over a confusing field, the sudden scroll back up to re-read something. You will usually see the same stumble two or three times, and that repetition is your answer.
  • Then a poll, if you're still unsure. Drop a one-question poll on the page — "Is there anything stopping you from continuing?" — and let real users tell you in their own words.

3. Fix the thing you actually saw

Now — and only now — you make a change. Not because a chart dipped, but because you watched people stumble on a specific thing. Then you go back to GA and check whether the funnel step improved. If it did, you understood the problem. If it didn't, you misread the recordings, and you go and watch again.

The bit that's really about communication

Underneath the tooling, this is a communication discipline, which is probably why it appeals to me. A website is a message the organisation is trying to send. GA tells you the message isn't landing. Hotjar tells you where the misunderstanding happens — the sentence that confuses, the ask that comes too soon, the button that doesn't look like a button.

A quick worked example

Let me make this concrete, because the abstract version is easy to nod at and hard to act on. Say your checkout funnel shows a brutal drop at the shipping step — sixty per cent of the people who reach it never move on. In Google Analytics that's all you know: a big red number at one step. You could spend a fortnight redesigning the entire checkout on a hunch.

Instead you go and look. The heatmap comes first, and on mobile it's damning: the "continue" button sits below the fold, and the scroll map shows a big chunk of visitors never reach it. They're not rejecting the step — they can't see the way forward. Then you watch ten recordings to be sure, and you notice something else: several people rage-click the order summary, clearly expecting it to be editable, then give up when nothing happens.

Now you have two specific, watchable problems instead of one vague number: a button people can't find, and a summary people mistake for a control. Both are small fixes — lift the button above the fold, make the summary visibly static. You ship them, and then you go back to GA to watch the funnel step recover. That's the loop working: a number found the wound, watching diagnosed it, the number confirmed the cure. No fortnight-long redesign of things that were never broken.

The numbers tell you a conversation is failing. Only watching people tells you what you actually said.

One honest caveat: recordings are seductive. It's easy to watch three dramatic sessions, fall in love with a theory, and forget those three aren't the whole population. Always let GA keep you honest about scale — is this friction hitting 40% of visitors or 4%? Qualitative shows you what's possible; quantitative tells you what's worth your time.

Use them in that loop — GA to find it, Hotjar to understand it, GA to confirm the fix — and you'll stop shipping confident redesigns that change nothing.