Most wearables now present a single daily figure describing how recovered the wearer is. That number is a weighted combination of several measurements, each with its own noise and its own meaning.
Heart rate variability carries most of the weight
The interval between consecutive heartbeats is not constant. It varies with breathing and with the balance between the two branches of the autonomic nervous system.
Greater variability generally indicates stronger parasympathetic influence, which is the state associated with rest and recovery rather than with mobilisation.
The measure is highly sensitive to posture, breathing rate, alcohol, illness and time of measurement, which is why devices capture it during sleep when conditions are most consistent.
Resting heart rate is the slow-moving companion
Lowest overnight heart rate is stable within a person and shifts predictably when something is wrong, rising with infection, dehydration, alcohol and accumulated training load.
Because it changes less from night to night than variability does, it acts as a stabiliser within the composite score, damping day-to-day noise.
A rise sustained across several nights is one of the more reliable signals a consumer device produces, and it frequently precedes any subjective sense of illness.
Temperature and sleep fill out the picture
Skin temperature is reported as a deviation from an individual baseline rather than an absolute value, because a wrist or finger sensor cannot measure core temperature.
Sleep contributions are usually duration plus an estimate of stage distribution, both inferred from movement and heart rate rather than measured directly.
Stage estimates are the weakest input, since distinguishing sleep stages properly requires brain activity recording that no consumer wearable performs.
The weighting is proprietary and arbitrary
How these inputs combine into one number is a commercial decision, and manufacturers do not publish their formulas or validate them against a shared standard.
Two devices worn on the same body on the same night will produce different scores, which tells you the composite is a product design rather than a physiological quantity.
Comparing scores between people or between brands is therefore meaningless, even where the underlying raw measurements agree reasonably well.
Baselines make the first weeks unreliable
Every input is scored relative to the wearer's own history, so a device needs several weeks of data before its comparisons carry information.
Scores produced during that learning period reflect an incomplete baseline, which is why early readings often swing widely for no apparent reason.
The useful signal is a multi-day trend rather than any single morning's figure, and a device that consistently reports poor recovery alongside symptoms is a reason to see a doctor rather than to train differently.