Constellations, patent pending

Audience Perception Maps · Reading the Data

What Heatmaps Actually Tell You

And what they don’t. Three tools share one word. Only one of them measures meaning, and even that one is raw material until you read the high scores against the low.

Quick answer

“Heatmap” means three different things depending on who you ask. A mouse-tracking heatmap (Hotjar) shows where cursors move on a live page. An eye-tracking heatmap (Tobii) shows where eyes land in the first instant, which is mostly survival-level reflex. A perception heatmap, the kind Audience Perception Maps produce, sits higher up: it shows what a concept actually looks like to people, captured by asking them to connect images to an idea and respond intuitively. Neither of the first two tells you what an image means to the person looking at it. And the map itself is the smallest part of the insight. It tells you what people liked and disliked. The read comes when you set the high-scoring images beside the low-scoring ones and follow the pattern between them.

Three maps, three completely different things

The word has been stretched over tools that have almost nothing in common. Marketers hear it and picture mouse-tracking. UX researchers picture eye-tracking. Both are useful, and neither measures what a perception map measures. If you are going to make creative decisions from a heatmap, you need to know which one you are looking at.

Mouse-tracking maps are good for usability. They show where people move and click on a live page. That is about the extent of it: a real, useful thing, and a narrow one.

Eye-tracking maps measure something more primal than people assume. The research is consistent, and brain scientists tend to agree: eye-tracking captures base, survival-level reactions. Where your eye goes first is driven by what evolution wired you to notice: danger, food, the body. It is a reflex. It tells you what grabs the eye in the first half-second. It does not tell you what the image means to the person looking at it.

Figure 1 · Three tools, one word

Mouse-tracking, eye-tracking and perception maps as three levels of processing Three blocks rising left to right against a vertical axis labelled level of processing. Mouse-tracking, Hotjar, where the cursor moves on a live page, tagged usability. Eye-tracking, Tobii, where the eye lands in the first half-second, tagged reflex. Perception map, Audience Perception Maps, what the concept looks like to people, tagged meaning and highlighted. LEVEL OF PROCESSING Mouse-tracking Hotjar where the cursor moves on a live page USABILITY Eye-tracking Tobii where the eye lands in the first half-second REFLEX Perception map Audience Perception Maps what the concept looks like to people MEANING
The three are not competing versions of one measurement. They sit at different heights. A cursor trail and a first-glance fixation are both real, and both stop well short of what an image means to the person looking at it.

Perception maps sit at a different level entirely. They do not measure reflex. They measure association.

What a perception heatmap actually measures

A perception test asks people to respond intuitively and emotionally to a concept. The prompt can be a brand, a value proposition, a feeling, a desired future, whatever idea you need to understand. Then it asks one simple thing: show us what that looks like. Which of these images feel like they represent the idea?

People connect a concept to a series of photos or images. It is intuitive, it is easy, and it feels almost subconscious. But it operates at a more thoughtful, more cerebral level than “where did my eyes go when you flashed pictures at me.” It is higher-level reasoning, captured quickly enough that people do not overthink it.

That is the distinction that matters. Eye-tracking tells you what is eye-catching. A perception map tells you what is meaningful. For brand work, meaning is the entire game.

The heatmap alone is the smallest part of the insight

Here is what most people miss the first time they see one of these maps: the heatmap by itself only gets you partway.

On its own, a perception heatmap tells a real but shallow story. People like this image. People do not like that one. Useful, but it is the opening, not the conclusion. The map alone is where most people stop, and stopping there leaves the best insight on the table.

The depth shows up when you arrange the results. Take the collection of high-scoring images and set it next to the collection of low-scoring images. Put what the audience connected with beside what it rejected. That is where the patterns emerge: the consistency, the connections, the through-line that says what the audience is responding to and why. Side by side, the data stops being a scoreboard and becomes a guide to what to do and what not to do.

Figure 2 · The map is the raw material

One perception map sorted into resonance, contested and resistance collections On the left, a board of sixteen image tiles, each carrying a scatter of cyan and purple response dots in no particular order. An arrow leads right to the same tiles sorted into three columns: resonance, mostly cyan, what the audience pulls toward; contested, an even mix, what the audience splits on; and resistance, mostly purple, what the audience pushes away. A bracket under the columns reads: read the two ends against each other, the middle clears in their light. THE MAP AS IT COMES BACK RESONANCE what the audience pulls toward CONTESTED what the audience splits on RESISTANCE what the audience pushes away Read the two ends against each other. The middle clears in their light.
Nothing on the right was not already on the left. The sort is the insight. Sixteen individual verdicts become two collections you can read against each other, and a third that only makes sense once you have.

Most people understand the map quickly when they see it. What takes a beat longer is realizing that the map is the raw material, not the finished read.

Resonance maps, resistance maps, and the contested middle

Two collections do the heavy lifting, and a third has turned out to be one of the most interesting.

Resonance maps are the images that scored high. Lots of green dots. These are the things people liked or associated with the prompt: what the audience is pulling toward.

Resistance maps are the images that scored negatively. Lots of red dots. These are the things people disliked or did not connect to the concept: what the audience is pushing away from.

Each is useful alone. Together they are the whole point. Seeing the resonance collection next to the resistance collection, what the audience embraced beside what it rejected, is where the software earns its keep. You are not guessing at direction anymore. You can see the boundary.

Then there is the contested middle: the images people split over, the ones that come back muddy, a mix of green and red. It is tempting to throw those out. Don’t. The contested pile often holds the most interesting information about an audience, because a split tells you something a consensus cannot. And those split images get much easier to interpret once you read them in the light of what clearly resonated and what clearly did not.

What a good map looks like vs a bad one

A bad map would look brown. A muddy wash, agreement nowhere, signal smeared across the whole board.

Here is the honest part: across dozens, arguably hundreds, of these tests, that has not happened. There is always a clear arrangement of green and red. There are tests with brown patches in the contested middle, but there is always a winner and a loser. There has yet to be a failed test, one that comes back as pure mud with nothing to read. It will probably happen eventually. It just has not yet, and that consistency is itself a finding: when you ask people to respond to a concept this way, they agree more than the “it’s all subjective” crowd expects.

Figure 3 · The failed test, and every real one

A hypothetical muddy heatmap next to the clustered pattern every test has produced Two boards of twenty-four tiles. The left board is washed in an even mix of cyan and purple at low opacity, labelled agreement nowhere, has not happened yet. The right board has cyan-dominant tiles on the left marked a winner, purple-dominant tiles on the right marked a loser, and a dashed outline around the two middle columns marked contested band. A FAILED TEST WOULD LOOK LIKE THIS WHAT EVERY TEST SO FAR HAS LOOKED LIKE contested band a winner a loser agreement nowhere — has not happened yet
The left board is a hypothesis. Across dozens, arguably hundreds, of tests it has never come back. The right board is the shape every test has taken: two clear ends, and a band in the middle that reads once you hold it against them.

A good map, then, looks like clear clusters. Defined green where the audience aligned, defined red where it recoiled, and a contested band in the middle that gets clearer the moment you set it against the two extremes. That is what testing a direction actually returns.

Which is the answer to what a heatmap tells you. Not where people looked. What they meant.

Frequently asked questions

What does a Constellations heatmap measure?

It measures perception: what a concept looks and feels like to an audience, by asking people to connect images to an idea and respond intuitively. It captures association and meaning, not where the cursor moved or where the eye landed first. The output is a map of resonance, resistance and the split between them.

How is it different from a Hotjar or Tobii heatmap?

Hotjar tracks mouse movement on a live page, which is usability. Tobii tracks eye movement, which is reflex-level attention in the first half-second. A perception map tracks what people connect to a concept, which is meaning. They operate at three different levels, and only the last one tells you what your creative communicates.

How do you read a perception heatmap?

Do not stop at the single map. Put the high-scoring images next to the low-scoring images and read the pattern between them. Resonance beside resistance is where the real direction shows up, and the contested middle becomes readable once you set it against those two ends.

What does a bad or inconclusive heatmap look like?

It would look brown: a muddy mix with no clear clusters. In practice these tests have always produced clear winners and losers, with any muddiness confined to a contested middle that clarifies once you compare it against what clearly resonated and what clearly did not. A fully failed test has not happened yet.