Reading a Heatmap Without Fooling Yourself

Why they are persuasive

A heatmap produces a picture, and a picture feels like evidence in a way a table of numbers does not.

It is also the only analytics output that non-technical people in a meeting will look at willingly, which is genuinely valuable.

The risk follows directly from that. Everybody can read it, so everybody forms a conclusion, and the conclusions usually go well beyond what the tool recorded.

The three kinds, which get conflated

• Click maps, showing where people tapped or clicked.
• Move maps, showing where the mouse pointer travelled.
• Scroll maps, showing how far down the page people got.

They answer different questions and have very different reliability. Presented in the same interface with the same colour scheme, they get discussed as though they are one thing.

Scroll maps are the most trustworthy, click maps are useful with care, and move maps are the ones that cause the most confident wrong conclusions.

Mouse movement is not attention

The assumption underneath move maps is that people look where they point.

That correlation exists and it is far weaker than the picture suggests. People rest the pointer wherever their hand happens to sit, park it in the middle while reading, use it to track a line of text, or leave it entirely still and read the whole page.

On a phone there is no pointer at all, so a move map for mobile traffic is either absent or is inferring attention from scrolling and touch, which is a different thing again.

Treat a move map as a rough hint and never as a statement about what somebody read.

False hotspots on click maps

The most common misreading, and it produces the wrong action.

A bright spot where nothing is clickable does not mean people love that element. It usually means they expected it to be a link and it was not.

That is a genuinely useful finding, and it is the opposite of the conclusion people draw. A hotspot on a non-clickable image or heading is a request, not an endorsement.

Other false hotspots come from a page where people tap repeatedly because nothing happened, which registers as enthusiasm and is actually frustration.

Sample size decides everything

A heatmap built from forty sessions is a picture of forty people, and it will look convincing regardless.

That is the trap: the visual does not communicate uncertainty. A hotspot from three clicks looks much like a hotspot from three hundred.

Check the session count before reading anything, and be sceptical below a few hundred for a click map.

Check the date range as well. A map covering a period that includes a campaign, a holiday, or a site change is averaging two different populations, and the result describes neither.

Mobile and desktop must be read separately

Combining them produces a map of a page that does not exist.

The layouts differ, elements sit in different places, and something at the top of a desktop page may be halfway down on a phone.

Averaged together, a hotspot appears at a position that is meaningless on either device.

Every tool separates these. Use the separation, and look at mobile first, since that is usually where most of the traffic and most of the problems are.

What they are genuinely good for

Three things, and they are worth the subscription on their own.

Finding out how far down a page people actually get, which is frequently much less far than assumed and changes where you put important content.

Finding elements people try to click that are not clickable.

Discovering that something you consider important is receiving no interaction at all, which is a clear signal even on a modest sample.

A worked example

An equipment supplier saw a strong hotspot on a product photograph and concluded the images were doing well, so they commissioned more photography.

The scroll map, which nobody had opened, showed that about seventy percent of visitors never reached the specification table below it.

The click hotspot was people tapping the photograph expecting it to enlarge. It did not.

They made the images expandable, moved the key specifications above the photograph, and shortened the introduction that had been pushing everything down.

Enquiries from product pages rose noticeably.

The original conclusion, more photography, would have cost several thousand and addressed nothing that was actually happening.

What they cannot tell you

Why anybody did anything, which is the question you usually have.

Whether the people clicking were the people you want, since a heatmap does not distinguish a serious buyer from somebody who arrived by mistake.

Whether somebody who left was satisfied or frustrated. A short session on a page answering a question quickly looks identical to a short session on a page that failed.

For any of those you need to watch somebody use the site, or ask them, and a heatmap is best used to decide which page is worth that attention.

The counter-case

There is a version of this scepticism that becomes an excuse to ignore the tool, and that is also a mistake.

A scroll map showing that two thirds of visitors never see your prices is actionable on its own, needs no further interpretation, and is not available from anywhere else.

Recordings of individual sessions, which most of these tools also provide, are more useful again for a small site, because ten watched sessions beat an aggregated picture of forty.

The discipline is not to distrust heatmaps but to keep each conclusion inside what was actually recorded, and to treat everything else as a question to investigate rather than an answer.

Reading one properly

1. Check the session count before anything else.
2. Check the date range for campaigns or site changes.
3. Split mobile from desktop and read mobile first.
4. Start with the scroll map, which is the most reliable.
5. Treat hotspots on non-links as requests, not approval.
6. Ignore move maps unless nothing else is available.
7. Use it to choose which page to watch somebody use.

Step seven is where the value is, because the heatmap narrows the search and something else answers the question.

Measuring actions rather than views is covered in events instead of pageviews.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.