PHYS-46 · Principle

Smart scale: body fat up 2% overnight. Real?

No. It measures resistance to a current and estimates fat by equation; a change can be missed by several points, so 2% is noise. Watch calories, sleep stages and VO2max are estimates.

How the body works, agreed across the reviews below; a study entry shows what one experiment measured.

well supported, with open questions namedApplies to every consumer deviceWell supported — what a sensor measures agrees with a reference far better than what an algorithm infers, and composite scores have not been formally validated (a living umbrella review, REF-1248; meta-analyses…

Practical Application

Smart scale: body fat up 2% overnight. Real? No. It measures resistance to a current and estimates fat by equation; a change can be missed by several points, so 2% is noise. Watch calories, sleep stages and VO2max are estimates.
What it means for training: name the reading's class first; act only on a measured reading on one device read as a trend, and give an estimate's error and ask (SYNTH-91).
What it does not mean: that wearables are worthless (a chest strap matched an ECG beat for beat, REF-1250), or that one brand's number can be compared with another's (REF-1251).

Mechanism

Measured: a chest strap tracked an ECG from rest to exhaustion (REF-1250); wrist optical heart rate was close at rest and running and read low on the bike and in the gym (REF-1249). Error comes from proprietary algorithms, fit, skin and constant firmware updates (REF-1248).
Estimated: watch VO2max is modelled from heart rate against running speed, or from resting heart rate, HRV and personal data, and missed one person by about ±10 mL/kg/min after runs (REF-1253); sleep is inferred from movement, heart rate and HRV, about 17 min high on total sleep (REF-1252), with stages 50-65% in agreement (REF-364); calories, probably from movement, wrist heart rate and GPS, read 17-29% high on two watches (REF-1254).
Body fat: bioimpedance gives fat-free mass and body water through population-specific equations, valid in people without fluid or electrolyte disturbance (Kyle 2004); 15 devices repeated themselves well but missed a laboratory model by −3.5 to +11.7 points on average (73 non-athletes, REF-1257).

Common misreading

"My watch VO2max dropped 2 points, so I am losing fitness"; "I burned 3 000 kcal, so I eat 3 000"; "40 min of deep sleep means poor recovery".

Backs

SYNTH-91 "Name the reading's class before using it." · SYNTH-91 "A new device starts a new baseline." · SYNTH-78 "The number on a watch or head unit is an estimate of the same kind and deserves the same caution"

Sources

Doherty et al. 2024 REF-1248 · Paper ↗
Molina-Garcia et al. 2022 REF-1253 · Paper ↗
Lee et al. 2025 REF-1252 · Paper ↗
Kyle et al. 2004 · Paper ↗