Smart scales are popular because they turn a daily weigh-in into a dashboard: body fat %, muscle, water, sometimes even “visceral fat” scores. Most of these numbers come from bioelectrical impedance analysis (BIA). The idea is simple: water with electrolytes conducts electricity well, while fat conducts poorly.

By measuring how much the current is resisted (impedance), the device estimates total body water, then converts that into fat-free mass and fat mass using prediction equations. This is scientifically grounded, but it also means the result is only as good as the assumptions the equation makes about hydration and body shape.

When BIA is compared to DXA, the headline finding is that agreement can be good for many people—especially for lean mass—yet bias appears at the extremes of body fat and body size. In a large healthy sample (591 adults), multifrequency BIA showed a strong overall correlation with DXA, but the average body-fat percentage from BIA was lower than DXA.

The key detail was pattern: BIA tended to overestimate body fat in lean individuals and underestimate body fat in people with higher body fat, with sex-specific cutoffs reported in the study. That means two people could step on the same scale and get “accurate enough” numbers, but someone very lean or more obese is more likely to see a consistent directional error rather than random noise.

Device design also matters. An eight-electrode, segmental, multi-frequency BIA system (the kind more similar to premium consumer devices or clinic-grade models) was compared with DXA in 132 adults across BMI categories. The results showed it was valid for normal-weight and overweight groups, but it overestimated percent body fat in the obese BMI group (about +3.4%).

The study also found the error increased as percent body fat increased, and waist circumference was the strongest predictor of systematic error—important because it suggests fat distribution (especially around the trunk) can distort BIA estimates. This fits a known limitation: many BIA models estimate the trunk less precisely than limbs, and trunk composition is exactly where many people store fat.

A larger middle-aged cohort (484 participants) evaluating direct segmental multi-frequency BIA also found excellent agreement for whole-body lean mass and strong agreement for fat mass and percent body fat, but with a nuance: Bland–Altman analysis showed narrower limits (less spread) for lean mass and relatively wider limits for fat mass and percent body fat. Segmental results were strongest in limbs and weaker in the trunk. In other words, smart scales and BIA devices tend to be best at estimating lean mass totals and less reliable when trying to precisely pin down fat mass—especially regionally.

Evidence in adolescents with obesity points to a similar story: BIA (Tanita MC-780) showed strong overall agreement with DXA for whole-body composition and extremely high reproducibility across repeated tests, yet it still overestimated body fat percentage and produced less reliable segmental estimates as body fat increased. This distinction matters for consumers: a device can be very consistent day-to-day (great for tracking change) while still being offset from the reference method (less ideal for absolute “true” body fat %).

Guideline-level reviews emphasize why standardization is crucial. ESPEN’s review notes that BIA can estimate fat-free mass and total body water in people without significant fluid or electrolyte abnormalities, but results depend on using appropriate equations and consistent procedures. It also flags hydration state as a major challenge, especially for more advanced segmental or multi-frequency approaches when body water distribution is altered.

Taken together, smart scales provide health data that is often directionally useful, but the most accurate interpretation is as a trend tool rather than a precise diagnostic instrument.