A smart scale can weigh you reasonably well while giving a much less dependable estimate of body fat. Those outputs come from different processes. Studies of consumer devices also show that producing the same number twice is a different achievement from measuring a person’s body composition accurately—or tracking how it changes.

For a reader looking at a body-fat percentage in an app, the useful question is therefore which claim has been tested: the weight reading, the estimate on one occasion, or the change between occasions.

The scale measures a signal, then estimates the composition

A conventional scale measures body weight. A body-composition scale adds bioelectrical impedance analysis, usually shortened to BIA: a small electrical current is used to measure how the body resists that current.

Withings’ explanation for Body+ describes a chain of estimates from impedance to total body water, then lean mass and fat mass. The displayed fat percentage is not a direct count of fat tissue. The calculation turns a measured electrical property into a model of the body.

That also explains why every number in the app is not a separate compartment that can be added together. Withings notes that water is already present in muscle and fat; adding water mass to the other composition figures would count overlapping material twice.

What the validation studies actually tested

A 2021 French hospital study compared three then-available scales with dual-energy X-ray absorptiometry, or DXA. The researchers tested different patient groups on each scale: 53 for Téfal Body Partner, 52 for Terraillon DietPack and 48 for Nokia Withings Body Cardio.

Weight measurements were relatively close to the reference, while body-composition estimates showed much larger discrepancies. Participants had been referred for body-composition assessment, often because of obesity or chronic conditions. This was not a head-to-head test in which every participant tried every scale, and it was not validation of today’s entire market.

A separate fifteen-device study, first published online in 2022, adds an important qualification. It tested 73 generally healthy adults against a laboratory four-compartment model that combined several measurements. Thirty-seven returned after 12–16 weeks, allowing the researchers to examine change over time.

Some models performed more usefully than others, including for tracking change. That finding argues against both blanket trust and blanket dismissal. It does not establish that every scale can accurately track an individual, or that results transfer unchanged between healthy volunteers, clinical populations and new hardware.

Three meanings of “accurate”

Repeatability

Does the device give a similar result when the measurement is repeated under similar conditions? A consistent offset can produce highly repeatable readings that remain far from a reference measurement.

Agreement on one occasion

How close is the estimate to the reference for this person? A study’s average result can conceal substantial individual errors. Strong correlation across people is also not the same as close agreement for each person.

Tracking change

Does the device capture a change in body composition rather than simply repeat its earlier estimate? The returning participants in the fifteen-device study addressed this separate question. A stable starting error does not automatically prove that later changes will be correct.

What to look for behind the percentage

Useful validation identifies the exact model, population, reference method and outcome tested. Evidence about weight cannot validate the fat estimate; evidence about immediate repeats cannot establish long-term tracking. The fifteen-device study used controlled measurement conditions, including attention to hydration and recent activity, so it cannot establish identical performance under all home conditions.

The practical distinction resembles the problem with estimating calories from a food photograph: a polished output can hide a chain of assumptions. For smart scales, the number of displayed decimal places tells you how the result is presented. It does not, by itself, tell you how close that result is to your body composition.