Your Heatmaps Are Lying to You

Photo by Pawel Czerwinski on Unsplash

It’s 2 PM on a Tuesday and your GA4 reports look flat. You’d love for something to jump out at you. Just one insight — any insight. Enter Hotjar or Microsoft Clarity. You start watching the session replays.

At first, it’s magic. You see a real person’s mouse zig-zagging across the screen. You see them rage-click on a button that isn’t a button. “Aha!” you shout to your empty office. “I’ve found something!” You feel like a digital detective.

But then you watch another. And another. After the fifth replay, the magic and your memory of the first watch start to fade. After the tenth, your eyes glaze over. You’ve got a notebook full of random observations, but no real, actionable truth. How many of these do you have to watch to actually find a pattern? Another couple handfuls? Dozens?

And that’s when the real danger kicks in: you start seeing what you want to see. You ignore the nine users who breezed through the checkout and focus on the one who struggled, because it confirms your pre-existing belief that the checkout is broken.

The immense popularity of these tools reveals a dirty little secret of the analytics world: we all know, deep down, that our standard, event-based tracking isn’t enough. The fact that we need these tools proves that the old way is broken. They’re a patch on a leaky tire. And spoiler alert: the solution isn’t a better patch.

The Right Instinct, The Wrong Tool

Let’s be clear: the instinct to use heatmaps is 100% correct. It comes from a frustration with what I call imperative analytics. That’s what I label the process of telling your analytics upfront exactly what’s worth measuring. Track acta_click here and a form_submit there. In the context of regression analysis, you’re putting everything else in the “residual error” box. That means the true contribution of any browser event not measured is going to splatter all over the stuff you are measuring in unpredictable ways.

Heatmaps and session replays are our desperate attempt to see the stuff that happens in between those predefined events. They are a gap-filler. The problem is, they trade one bias for another. You go from the bias of pre-selected events to the self-selection bias of only seeing the patterns you were already looking for.

Watching individual session replays to understand your user base is like trying to understand a wedding reception by looking at a handful of random photos. In one photo, Uncle Jerry is spilling wine. In another, two people are having a great conversation. In a third, someone is staring at the wall.

What does it all mean? You have no idea. You have anecdotes, not insights. You’re missing the flow, the energy, the overall shape of the party.

Letting the Math Search Out the Beauty

“Okay,” you’re thinking, “but how do I look at a blueprint of millions of events without going crazy? A bigger dataset just sounds like a bigger headache.”

You don’t look at it. Not manually, anyway. You let the math do the heavy lifting for you. Topological Data Analysis (TDA) seems either 1) intimidating or 2) something invented so consultants can sound smart and find work to do. Don’t let the name scare you. And TDA is used everywhere except marketing. Consider the lilies of the field; yep, botanists, too.

Think of it like using the zoom function on a blueprint. As you zoom in and out, some patterns are fleeting — just a few people who happened to stand together for a moment. That’s noise. But some patterns persist. The dense crowd at the bar is a real cluster whether you’re looking from a hundred feet up or ten feet away. The empty space at the photo booth is still empty at every zoom level.

TDA is designed specifically to find these features that persist across a wide range of scales, flagging them as mathematically significant. It automatically separates the real, structural patterns in user behavior from the anecdotal noise.

So, instead of you watching one session replay and seeing a user bounce between the pricing page and the enterprise features list, the math does the work. It analyzes every journey and comes back to you, saying: “We have a statistically significant ‘PolicyCheck Loop’. Hundreds of users are caught in this exact pattern, indicating they are actively building a business case but can’t find the information they need to seal the deal”.

That’s the difference between an anecdote you hope is a trend and a systemic issue you know is worth fixing.

Photo by Bernd 📷 Dittrich on Unsplash

A More Natural Science

Let’s go back to those lilies for a second. A botanist uses TDA to see the underlying structural rules of a plant — the branching patterns of its veins, the way its petals cluster. They aren’t just looking at one leaf and guessing how that might’ve evolved; they are mapping the fundamental geometry of its design to understand the system as a whole. They’re looking for the persistent, architectural truths that make a lily a lily.

We could be doing the exact same thing but you pivot table too much.

Your user data is an ecosystem. It’s not a collection of random clicks any more than a plant is a random collection of cells. It has a structure. It has a shape. The behavioral loops and voids that TDA reveals are the vein structures of your customers’ intent. They show you how information flows, where it gets stuck, and where it nourishes growth.

For too long, marketing has tried to operate like a tourist, taking random snapshots (heatmaps, session replays) and hoping they tell a story. This is a chance to start operating like a botanist — a natural scientist observing the ecosystem, identifying its core patterns, and making decisions based on the structural truths of the system itself.

The instinct that drove us all to Hotjar was the right one. We knew there was more to the story. We knew that staring at dashboards wasn’t enough. We just need to trade in our snapshots for LiDAR. Stop guessing, and start seeing the shape of things.