Liquidation heatmaps show bands of colour at price levels, implying that a great deal of forced selling waits there. They are genuinely useful and they are widely misread, because most people treat them as observation when they are inference.
What is actually known
Venues publish liquidations after they happen, usually with size and price. That part is observed. Aggregate open interest is observed too.
What is not observed is where individual positions sit or what leverage each uses. The map is built by assuming a distribution of entry prices and leverage levels across the known open interest, and computing where those hypothetical positions would be liquidated.
Why the assumptions matter
Change the assumed leverage mix and the bands move. A map built assuming heavy use of high leverage puts clusters close to price; one assuming conservative leverage puts them far away. Both are drawn with the same confidence.
This is why two providers produce different maps from the same public data, and why the absolute values are not comparable across providers.
What holds up anyway
The relative picture is more robust than the levels. Substantially more estimated exposure below price than above tells you the asymmetry, and that conclusion survives most reasonable changes to the assumptions.
The direction of change over time is also informative. A cluster growing session over session means leverage is accumulating in a region, regardless of whether the exact level is right.
How to use it
As a map of where a move could accelerate, not as a target. Price does not seek liquidations, and treating a band as a magnet is a good way to hold a losing position waiting for a level that has no obligation to arrive.
Confirm it afterwards with open interest: a real flush shows up as a drop, and no drop means the cluster is still sitting there.