What Smart-Scale Body-Fat Readings Can and Can't Tell You

How does a smart scale estimate body fat?

It does not measure fat at all. The method is bioelectrical impedance analysis (BIA): electrodes under your feet, and on some models in hand grips, pass a weak current through your body and record how strongly it resists. Body water carries the current and fat holds little of it, so the scale feeds that resistance, plus the height, age and sex you entered, into a manufacturer's equation that predicts fat-free mass and total body water. Fat mass is the remainder of your weight. A Clinical Nutrition review states the premise: BIA can determine fat-free mass and total body water in people without significant fluid or electrolyte abnormalities, using equations suited to their population and age.

Two design details shape the result. A foot-to-foot scale sends current through the legs and lower trunk only, while models with hand grips and eight electrodes also cross the arms and upper trunk. And every equation was built on a reference group, so accuracy depends on how closely you resemble it in age, weight status and ethnicity. The authors of a 2008 comparison of consumer devices asked manufacturers to disclose the reference population and method behind each equation; a 2023 team testing 15 devices still found little accessible information on how they were developed and validated. The "smart" part, the app and Wi-Fi, adds storage and charts, not a different measurement.

How accurate is smart-scale body fat compared with a DXA scan?

For an individual, less accurate than the display implies. Three comparisons show the pattern.

• A 2008 study of four single-frequency consumer devices in 106 adults, with MRI and DXA as references, found small average differences: under about 1.5 percentage points for three devices, about 3 for the fourth. Individual differences were wide. The 95% limits of agreement ran from about 9 points too low to 15 too high for the least accurate foot-to-foot unit, and about 5 too low to 7 too high for the hand-grip model.
• In 2023, 15 devices (14 of them consumer-grade) were compared with a four-compartment model in 73 healthy adults. Average error ranged from −3.5 to +11.7 points and standard errors from 3.1 to 7.5. Only five, one of them the research-grade unit, were statistically equivalent to the reference within ±2 points.
• Against DXA in 929 adults, a foot-to-foot scale had a root-mean-square error of 5.6 points in men and 4.6 in women and failed equivalence tests at ±2.5 and ±5 points.

The error has a shape. Many devices over-read lean people and under-read people with more body fat: all three foot-to-foot units in 2008 and six of the 15 devices in 2023, whose authors advise anyone far from the population average to be more skeptical. Design matters too. In both studies, models with hand electrodes generally outperformed foot-to-foot ones, and the 2008 paper adds that measuring at several fixed frequencies may confer no advantage over a single one.

DXA is the usual yardstick, but it is not truth. Questions have long been raised about how hydration, software version, hardware and the population studied affect its estimates, and a perspective review argues that BIA's errors are not dissimilar to those of so-called gold-standard methods. When a scale and a DXA report disagree, both contribute to the gap. For how validation studies are read across other wearables, see our Wearable Accuracy Guide.

Weight is a different matter. In a 2020 test of a multi-frequency Wi-Fi scale against air-displacement plethysmography in 20 young adults, mean absolute percent error was 0.15% for body mass and 25.8% for fat mass; the authors advised caution with the fat figures.

Why does the reading change between morning and evening?

Because the scale is reading water. Meals lower impedance for two to four hours, and the equation interprets that as less fat. In a study that measured 18 healthy adults 18 times over 24 hours, the calculated body-fat percentage varied by 8.8% in women and 9.9% in men between the highest and lowest readings of the day. Exercise does the same: after 30 minutes of cycling, 74 adults showed lower estimated body fat, by 0.3 to 0.5 on a leg-to-leg analyzer and 0.8 to 1.4 on a segmental one, with larger drops after harder exercise. Neither changes fat mass; the shift is in the signal.

If my scale shows the same number every time, is it accurate?

Not necessarily, because repeatability and accuracy are separate properties. In the 2023 comparison, every device's precision error on back-to-back readings was under 0.5 points, yet standard errors against the reference ran from 3.1 to 7.5. Five foot-to-foot models returned readings within 0.1 point of each other on immediate retest, which the authors called suspiciously consistent, suspecting software that holds repeat readings from one user account close together. A second step on the scale may not be a second measurement.

Can a smart scale track body-fat changes over time?

Sometimes, depending on the model and the size of the change. The best case is a large change. In a 12-week weight-loss study of 58 women who lost about 10 kg, 7.6 kg of it fat, a leg-to-leg scale estimated the fat lost with an average error of about 0.5 kg, though the 95% limits spanned roughly ±3 kg.

The ordinary case is harder. In the 2023 study, 37 adults returned 12 to 16 weeks later; the reference model showed their body fat had moved by only 0.2 points on average (range −5.0 to +7.5), and the devices' standard errors for that change were 1.7 to 2.6 points. Nine of the 15 matched the reference's average change within ±1 point, yet the authors told consumers to be wary, while noting that some models may be useful. The ESPEN clinical guideline is similarly careful: follow-up by BIA is possible at a BMI of 16 to 34 with normal hydration, but must be interpreted with caution.

Which readings on the screen deserve the least trust?

• Visceral fat. In the 2008 study the devices' visceral-fat level agreed poorly with MRI, and the authors called it only a qualitative index.
• Muscle mass. The figure depends on what the manufacturer's equation reports. In that study one device reported bone-free lean mass, judged against DXA, while two reported skeletal muscle. In the same participants, MRI-measured muscle averaged 24 kg and DXA lean mass 51 kg, so one brand's "muscle" is not another's.
• Athlete modes. The 2023 team entered the lowest activity setting on every device that asked for one and listed athlete modes as an open research question, so that comparison did not test them.

Who should be most careful with these numbers?

• People far from the average. The ESPEN guideline did not recommend routine clinical use of BIA at extremes of BMI or with abnormal hydration, and the 2008 paper lists renal or heart failure and severe obesity as conditions that break the constant-hydration assumption. In a study of 5,740 nutrition-clinic patients with a mean BMI of 37.7, aleg-to-leg scale over-read fat mass against DXA with wide limits of agreement, and in the 72 women who had gastric bypass it overestimated the change in fat mass at 3 and 12 months.
• Athletes. In 33 male US college athletes, a foot-to-foot scale had 95% limits of about ±5.7 points and error that shifted with leanness. The authors judged it unsuitable for an individual's body-fat percentage, though acceptable for group estimates.
• People with implanted cardiac devices. A 2023 simulation and bench study of smart scales, watches and rings found simulated interference above ISO 14117 limits, lower for scales than for watches, and did not recommend these devices for people with implants. A 2025 review of six clinical-grade studies (531 patients) found no clinically significant interference but cautioned against extrapolating to consumer-grade devices. Ask your cardiology team before stepping on one.

Does a body-fat percentage tell you whether you are healthy?

Not by itself. When researchers proposed percent-fat ranges in 2000, they noted that no accepted published ranges existed and tied theirs to BMI categories. The 2025 Lancet Diabetes & Endocrinology Commission on clinical obesity says BMI alone can both underestimate and overestimate adiposity, and that excess body fat should be confirmed with at least one waist-based measure alongside BMI or, where available, a direct measurement such as DXA or bioimpedance, using validated methods and cutoffs appropriate to age, sex and ethnicity. In its framework, calling excess fat an illness requires signs of reduced organ function or limits on daily activities, not a percentage alone. A scale reading can start a conversation with a doctor; it cannot settle one.

How do you get the most useful readings?

The 2023 authors suggest three habits: measure each morning after an overnight fast from food and fluid, after using the toilet and in minimal clothing; track a multi-day average rather than single readings; and use the same device for tracking over time. Beyond that:

• Skip the scale after workouts and meals; the studies above show how far each moves the estimate.
• Treat a change of a point or two as noise. Under laboratory conditions with immediate retesting, the 2023 authors called a change of at least 1.4 points real on any device they tested, and day-to-day biology would push that higher. Judge direction over several weeks.
• Cross-check with measures that do not rely on an electrical signal: waist circumference, photos, strength numbers, how clothes fit.
• If watching the number affects how you eat or train, stop checking it and talk to a health professional.