Write down a question before you open the tool, then read one page and one device at a time.
A heatmap tells you where to look, never why: the cause is in five to ten recordings. Every finding then ends as a fix or a hypothesis to test.
The benchmark
At an audit debrief, a founder told us:
"We struggle a bit to analyze [...] because sometimes there are so many recordings, we don't really understand them."
I told her there was no miracle recipe:
"Grab some popcorn, watch at fast speed, and as soon as you see something a bit odd, hit pause."
I had watched dozens for her audit, and it's not a thrilling exercise, even less when you don't know what you're looking for.
A map is read against a baseline. In a desktop eye-tracking study published in 2018, the Nielsen Norman Group found that 57% of viewing time is spent above the fold.
A hot top and a cold bottom therefore describe normal behavior, and the insight lies in the gap. On the mobile product page of a fashion store that we audited, 43% of new visits reached the buy button and 21% reached the description.
The 21% looked normal to us for a description placed below the button. With 43%, more than one new visitor in two left without seeing how to buy.
In Clarity, the scroll map gives this depth in 5% increments: note where half of visits stop and what sits just below.
The rule
Separate the bug from the friction.
On that same founder's store, neither the title nor the text of the products was clickable in the collection, where only the image led to the product page. I saw six or seven sessions where the person tapped the wrong spot, tried again, then left, including one arriving from Instagram, possibly through an ad.
A defect like this gets fixed without a test, and everything else becomes a hypothesis to write like this: "On [page], [element] causes [signal] among [segment]. If we [change], [metric] should increase."
If you don't have enough traffic for a test, the decision is made in a meeting, using the body of evidence described in the guide on optimizing without A/B testing.
The steps
The example uses Microsoft Clarity on Shopify.
- The frame. Pick one question, one URL or a template of consistent length, and 14 to 28 days with no promotion or template change. Clarity keeps recordings for 30 days.
- The funnel. In the dashboard, the Purchases tile isolates the sessions that bought. On Shopify Plus, start with the Checkout Abandonment tile, which breaks abandonments down by contact details, shipping, payment, and "other."
- The device. Read mobile and desktop separately. The fold shown by Clarity is an average: check the page on a small phone.
- The recordings. Work by elimination with the filters: first dead clicks and JavaScript errors, where bugs hide, then high-intent sessions that didn't buy. A rage click points to a spot and leaves the cause open: a dead link or a carousel being used normally.
- The decision. For each finding seen in at least five distinct sessions, write the fix or the hypothesis, with the URL and the date.
The trap
Since October 31, 2025, Clarity requires a valid consent signal for visits from the European Economic Area, the United Kingdom, and Switzerland.
When this signal is missing, or the person declines, the tool keeps collecting without cookies, and every page view becomes a separate session. These visits produce more recordings, all of a single page, and funnels showing total drop-off at every step.
Volumes go up while journeys disappear.
Clarity detects the choice recorded by Shopify's native cookie banner. With a third-party banner, check that it passes this choice to Shopify.
From this, we derive a check: accept cookies on your store, visit three pages, then find your visit in the recordings.
If it shows up as three one-page sessions instead of one, have the banner checked before any other analysis.
The exercise
On Monday, block thirty minutes for a single template, on mobile, with the dead-click filter and ten recordings at fast speed. The case described above would produce this, to copy and then rewrite with your own finding for your leadership:
"In the collection, the product title isn't clickable. I saw it in six sessions this month. It's an implementation fix to ship this week, with no test."
