Open a page's analytics and you see, say, four thousand views. Open its heatmap and the click count is a fraction of that. Nothing is broken. The two are measuring different things, on purpose.
They start from different data
Page views are counted at ingestion: one row per load, every time.
A heatmap is drawn from interaction capture: the clicks and scrolling recorded while somebody is on the page. Session recordings are sampled at a rate you set per site; the clicks and scrolling behind a heatmap are not.
Four reasons the numbers differ
A heatmap covers one kind of device. A phone and a desktop lay the same page out differently, so their clicks cannot share a map, and each heatmap shows the kind you pick. Your analytics counts every device in one figure.
Some visitors opt out. A site can be set to honour Global Privacy Control, and where it does, a browser sending that signal is excluded from interaction capture while still being counted as a visit. That is the correct outcome and it makes the two numbers differ.
Parts of the page can be blocked. Masking rules and blocked selectors keep chosen regions out of what is captured. A section nobody appears to touch may simply be one you excluded.
Not every visit has interactions to record. A visitor who lands, reads and leaves without clicking contributes a page view and almost nothing to a click map.
Which to trust for what
Use analytics for how many: visits, entry pages, conversions, anything you would put in a report.
Use the heatmap for where: what people reach for, how far down they get, which of two calls to action wins. It answers a question counting cannot, and it is the wrong instrument for arithmetic.
The failure worth avoiding is quoting a heatmap's totals as traffic figures. They were never meant to be that, and they only ever cover one kind of device.
When they disagree in the other direction
If the heatmap shows clicks on a page your analytics says nobody visited, look at your exclusions. Separate tools rarely share a filter, and an exclusion added to one and not the other is the usual explanation. In Skomi one set of exclusions covers both, so check whether the report itself is filtered.
The two drawings answer different questions, and the glossary keeps them apart: a click map shows where people pressed, an attention map shows where they stayed, and scroll depth is neither.
Related: when you need session recordings, what bounce rate actually measures, and Behavior.