Heatmaps and session recordings: how to use them

Steps for starting from a funnel question, finding the place on a heatmap, looking for the reason in recordings, masking personal data and turning an observation into a hypothesis to test.

· 13 min readData

Short answer

A heatmap gathers the click and scroll behaviour of many visitors on one page; a session recording replays one visitor’s session. Neither proves the reason, both suggest a hypothesis. Take a question from the funnel, let the map show the place, let recordings suggest the reason, then test the result.

Picture this: on a services site, plenty of people start the quote form and few send it. In the meeting one person says the form is too long and another says nobody notices the button; neither has a number, only an opinion. Heatmaps and session recordings help exactly here: they let you see where visitors click on a page, how far they scroll and where they get stuck. In this article we first explain what each one shows and what it cannot show, then how to start from a question and arrive at a hypothesis.

We use Microsoft Clarity’s own documentation as the worked example, because it is public. Everything we quote is what Clarity says about itself; other tools may name things, set limits and offer settings differently, which is why we do not recommend a tool. The interface names in the text (Heatmaps, Click maps, Masking mode and so on) are the names in Clarity’s English interface. We covered finding the loss in the funnel in how to do a GA4 funnel analysis and the general framework in how to increase conversion rate; here we start where they stop.

What heatmaps and session recordings show

According to Clarity’s documentation, a heatmap is a visualisation that makes it easier to analyse aggregated information about how users interact with a website. Clarity collects clicks and scrolls on desktop, mobile and tablet and generates the map itself; the map combines data from many page views, according to the filters you choose. A session recording is one visitor’s session. The documentation says sessions are not real screen videos but step-by-step visual reconstructions built from the page’s HTML and the user’s actions. Our summary: a map shows the average of many people, a recording shows one person’s story. There is also a short definition in the glossary.

In the table below, the first two columns are the map types in Clarity’s documentation; the third column is our interpretation.

Map type (Clarity’s name) What the documentation says it shows What it cannot say on its own (our interpretation)
Click maps Where users click on the page; the types are all clicks, dead clicks, rage clicks, error clicks, first clicks and last clicks Why they clicked, and whether the clicker found what they were looking for
Scroll maps How far down the page users scroll, and the average fold Whether the content at that depth was read
Area maps Total clicks on all elements in a selected area What happens outside the area; choosing the area yourself also adds a bias
Attention maps Time spent on each part of the page Whether the time came from interest or from confusion
Conversion maps In e-commerce, which clicks lead to a purchase The effect of price or a campaign on sales

The list in Clarity’s documentation has no mouse-move map. If you see such a type in another tool’s heatmap, read what it measures in that tool’s own documentation.

We will go in five steps; the figure below summarises the order.

Heatmaps and session recordings in five steps: funnel question, map, recording, hypothesis, test 1 Funnel question Page, device, date range 2 Heatmap Place: clicks and scrolls 3 Session recording Reason: shared snags in recordings 4 Hypothesis Observation, cause, change, measure 5 Test A/B test or a fix Heatmaps and session recordings in five steps: funnel question, map, recording, hypothesis, test 1 Funnel question Page, device, date range 2 Heatmap Place: clicks and scrolls 3 Session recording Reason: shared snags in recordings 4 Hypothesis Observation, cause, change, measure 5 Test A/B test or a fix
Figure: The method starts with a question from the funnel, looks for the place on a map and the reason in recordings, and ends with a hypothesis and then a test.

1. Take the question from the funnel

Do not open a heatmap to “see what comes up”; without a question, every coloured area turns into a story. The question comes from the biggest loss step you found in the funnel: “On mobile, in which field do people who start the quote form leave?” or “Why don’t people who add to the basket go on to checkout?” (for the checkout step, see how to reduce cart abandonment). Add three things to the question: which page, which device, which date range. These are our recommendations; without all three, you read the map from the average of all traffic, and an average often describes no real visitor.

The question in one sentence might read: “At what point do [device] visitors leave while trying to [do the action] on [the page]?” If you cannot write the sentence, it is not yet time to open a map.

2. Find the place with a heatmap

In Clarity, you choose the page address and filters in the Heatmaps section. According to the documentation, with the URL parameters option on you see the map for the page matching the parameters exactly, and with it off you see the map for addresses that start with the value you gave. So you merge or separate a page’s campaign-parameter addresses on purpose. Then look at three things:

  • Click maps. The documentation says you can use this map to see whether your CTAs are getting any attention. You can open the dead clicks and rage clicks types separately; both are explained below.
  • Scroll maps. According to the documentation, they show what percentage of users scroll to a given part of the page, and the average fold. If your button or form sits far below the average fold, look at how many people reach it.
  • Device choice. Look at desktop, tablet and mobile separately; open whichever device your funnel question names.

What comes out of the map is a place marker: “this area gets many clicks but nothing happens when you click” or “the form sits where most people do not scroll”. Leave the question of why to the next step.

How many sessions before you read anything into it

Clarity’s documentation says there is no minimum traffic limit for generating a heatmap and that the map arrives within seconds. That means the tool draws a coloured map even from two visits, and a map being drawn does not mean you can draw conclusions from it. We did not see a threshold in the documentation saying “this many sessions is enough”, and we are not inventing one. Sample size can be calculated for A/B tests (see how to run an A/B test); we do not use a ready-made formula of that kind for maps.

Instead we suggest this rule (our rule):

  • Write down the page, the segment and the period before you look at the map, and do not change them after you have looked.
  • Add the number of page views the map is based on to your notes. As you narrow the segment (for example mobile only) that number shrinks; the smaller it gets, write “doubt” instead of “observation” in your interpretation.
  • Do not form a hypothesis before you have seen the same pattern a second time in a separate period.

3. Look for the reason in recordings

The map showed the place; now you watch what people do at that point. According to Clarity’s documentation, a session ends when the user has been inactive for 30 minutes, and the session card carries details such as entry and exit pages, referral URL, number of clicks and device type. Recordings are kept for 30 days; those added to favourites stay for up to 9 months. There is a “skip inactivity” option for skipping idle stretches.

To avoid drowning in the recording list, we suggest this routine (all of it is our suggestion):

  1. Narrow the list to your question’s segment with Filters and Segments: page, device, period.
  2. Start with those who did not complete the goal and those that carry a signal; for comparison, also watch a few who did complete it.
  3. Take one line of notes per recording: what they were trying to do, where they got stuck, what they did.
  4. Write the stopping rule before you start watching: stop when the same observation repeats and no new pattern appears. We do not give a fixed number of recordings.
  5. Add an important recording to favourites or note it down; after 30 days Clarity keeps only a small sample.
  6. Write a finding from a single recording as just one line of findings. Technical faults such as a button that does not work are confirmed by trying it yourself and do not need a test.

How Clarity defines the signals

Clarity’s Semantic metrics page defines the flags that mark sessions like this:

  • Rage clicks: the user clicking several times in rapid succession in a clustered area. The documentation says this can be a strong indicator of frustration where users expect an element to be interactive.
  • Dead clicks: a click on an element that gets no feedback within a reasonable time; the visual state does not change and there is no navigation away from the page. According to the documentation, it can point to a broken element, a high-latency request or misleading UX.
  • Excessive scrolling: vertical scrolling in a session that is higher than the expected average; as possible causes the documentation lists inaccessible or frustrating UX, poor content relevance or discovery.
  • Quick backs: the user going to a new page and returning to the previous one in less time than a defined threshold.
  • JavaScript errors and click errors: errors in the code running in the browser; when an error follows a click, the session is tagged automatically.

The documentation gives no numeric threshold for any of them, and its wording says “can”. Our reading: a flag is a way of choosing which recording to open, not a diagnosis.

4. Turn the observation into a hypothesis

An observation says what you saw: “On mobile, on the checkout page, people click repeatedly on the area under the delivery line and nothing happens.” A hypothesis adds a cause and a change: “People click this area because it looks like a link. If we turn the delivery information into a clickable explanation, the share of people who complete checkout will rise.” We explained how to write the four parts (observation, cause, change, measure) in the conversion rate article. One extra warning is enough here: you infer the cause from the recording, and the recording does not prove it; write the hypothesis as “we assume it works like this”.

If the observation is on a form page, the page order and form rules in how to build a landing page will also help.

5. Test the change

If the finding is a fault, you fix it. If not, you test the change; the detail is in how to run an A/B test. If traffic is not enough for a test, you can make the change and compare the maps from before and after; Clarity’s documentation says it has a compare feature for this. But because visitors are not randomly split over the same period, this does not count as a test; write the result as “tried, it looked like this”. This order matches the one written on Detmore’s data and CRO service page: funnel data and recordings show where the biggest drop-off is, and then one change is tested with a large enough sample.

Recordings contain the behaviour of real people; look at two things before you switch the tool on.

Masking. According to Clarity’s documentation, sensitive content is masked by default and masked content is not uploaded to Clarity. The setting is under Settings, in the Masking section, and there are three modes: Strict masks all content. Balanced (the default) masks only sensitive content, and the documentation counts numbers and email addresses as sensitive. Relaxed masks no content, but input boxes and dropdowns are masked in every mode. You can also mask specific elements with a CSS selector or the data-clarity-mask attribute. Note three details: only a project’s administrators can change the mode. A change applies to new recordings, is not applied retroactively and can take up to an hour to show. Content in style sheets and style tags is not masked, so do not put anything sensitive in CSS. Our suggestion: once the tool is on, try your own form, checkout and account pages from your own device and watch the recording to see what shows.

Consent. According to Clarity’s Consent management page, from 31 October 2025 the tool looks for a valid consent signal on visits from the European Economic Area, the United Kingdom and Switzerland. Without consent no cookies are placed in those regions, and the functions of session recordings and funnels may be limited. The page does not say what to do for visitors from Türkiye; that does not mean consent is not needed. Which law applies, what the privacy notice should say and how the recordings are assessed legally are questions for a lawyer; settle them with your own. For the cookie notice and the technical setup of consent mode, see how to set up cookie consent and consent mode; that article cites the KVKK cookie guide as its source. One conclusion is ours: if you have recordings only of people who gave consent, the people you watch may be a small and selected part of all visitors.

The limits of the method

  • What you see and what was lived may differ. Recordings are reconstructed images; they may not match the visitor’s screen exactly (our inference).
  • It does not show who never came. Someone who never reaches the page, or who decides by phone or in a shop, is not on the map.
  • Data ages when the page changes. Do not read data from before and after a design update on the same map.

Common mistakes

  • Assuming the most clicked element is the most useful. The menu or logo is always hot; what matters is the area that feeds the goal action in the funnel.
  • Watching only flagged sessions. People who leave quietly carry no flag.
  • Presenting a single recording in a meeting as a rule. “I saw it in one recording” is a line of findings; wait for the same pattern to repeat before deciding.

Common questions

How many sessions should I watch?

We did not see such a number in the documentation, and we do not give one. Write down in advance which segment you will watch and when you will stop; stop when the same observation repeats and no new pattern appears.

Can you see what visitors type in recordings?

According to Clarity’s documentation, input boxes and dropdowns are masked in every masking mode. Still, test the setting on your own site: make a recording with fake data on the form, checkout and account pages and look at what shows.

Does a heatmap replace an A/B test?

No. A map and recordings give hints about the cause; a randomly split test tells you whether a change worked. If traffic is not enough for a test, it is reasonable to go on with maps and recordings first, but do not write the result up as a test.

One thing you can do today: if the tool is installed on your site, fill in your quote or checkout form from your own device with fake data. Then watch the recording of that session and see whether the values you typed are visible. If they are, the next step is the masking setting; if they are not, pick the funnel step with the biggest loss, write its question in one sentence and open the map only after that.

Sources

This article is general information, not legal, financial or technical advice. Platforms, their rules and the law change over time, so some details may have changed since 14 November 2026 or contain small errors. Before you make a decision for your business, check the current official sources or write to us.

Tell us about
your project.

Fill in the form or write to us directly, and we will agree on a day to talk.

Your first meeting is with Deniz Sezer, founder of Detmore.

First meeting free · Reply within one working day · Written proposal in 2 to 3 days · NDA ready

What do you need? pick as many as you like

Name, email and phone are required.

Contact

We reply within one working day.