A heatmap is a compression. Compression throws information away. That is not a moral failure; it is a reason to refuse the picture as a finding.
In class we return any ticket that cites a heatmap without a sampling note. How many sessions? Which devices? Which language rendering? Was the blob produced by a sticky header that people were not even aiming at?
Thai type changes the picture
Line wrapping in Thai can move a tap target by a full row between desktop and a mid-range Android. A heatmap generated mostly from desktop sessions will “prove” that a button is well placed while mobile users miss it. If your vendor does not split device classes, the blob is a tourist map.
A bright region tells you where pixels were busy. It does not tell you whether the person understood the next action.
What we ask students to attach
Date window, sample size, device split, language, and one sentence on what the heatmap cannot show (hesitation, backtracks, recovery after error). Then, if they still want the picture in the brief, they may include it as illustration, not as proof.
Heatmaps remain useful for spotting accidental tap magnets and for showing a stakeholder that attention never reached a fee disclosure. They are a poor spine for UX Friction Analytics. The spine is timestamps plus a named family plus a limitation you are willing to read aloud.