Are body fat scales accurate?
DEXA vs. smart scale

Most arguments about body-composition devices go wrong in the first sentence, because they use one word for two properties. Accuracy is how close a measurement sits to the true value. Precision is how closely repeated measurements agree with each other. A bathroom scale that reads four pounds light every single time is precise and wrong.

Separate the two and the comparison stops being a contest. DEXA is the accurate instrument you use rarely; the scale is the approximate one you use every day. And if the scale's error is largely a stable bias (the same offset in the same direction each time), subtracting one reading from another cancels most of it. The change it reports is worth far more than the value it reports. That is the whole case for owning a device you have just been told is inaccurate, and it is the principle Bild Health's trend view is built on.

What each instrument physically does

DEXA passes two X-ray energies through the body and works backwards from their differential absorption to how much bone, fat, and lean soft tissue lay in the path. A body-composition scale runs bioelectrical impedance analysis: it sends a small alternating current between electrodes in its surface, measures the body's resistance to it, and feeds that resistance into a regression equation (one fitted against reference measurements in a sample population) to estimate total body water and, from it, fat-free mass. Fat mass is what follows, by subtraction from body weight.

Two things follow. The scale's answer depends on how closely you resemble the population its equation was fitted on, which nobody discloses and you cannot check. And because the signal is electrical and body water conducts, anything that shifts your fluid balance shifts the estimate: hydration, a recent workout, a salty meal, alcohol, time of day. This is measurable: in a controlled test, drinking two litres of water pushed impedance body-fat readouts up by roughly one to two percentage points within an hour. Hydration alone moves the number by more than a week of real change.

Which reading answers which question

All of it comes down to routing. Given a thing you want to know, which instrument do you look at, and which one do you ignore?

Getting a reliable trend out of an unreliable number

If the scale's absolute value is off but its bias is stable, everything depends on keeping the conditions stable too, because every condition you let vary gets read as body composition. The protocol that makes the trend trustworthy costs nothing:

• Same time every day: first thing in the morning, after the bathroom, before food, coffee, or water.
• Same state: not right after training, a sauna, alcohol, or an unusually salty evening; each one moves fluid, and fluid moves the reading.
• Same device, bare feet, same spot on a hard floor. Comparing across devices measures the devices, not you.
• Read the line, never the day. A single reading is hydration wearing a body-fat costume; two weeks of readings is a signal.

Held to that, the change-tracking result above is on your side: most tested devices measured twelve-week shifts about as well as the reference method. Let the conditions drift and you are charting your fluid balance with great confidence.

Are RENPHO and Withings scales accurate? Reading the vendors' own numbers

The two brands people actually search for make instructively different claims. RENPHOpublishes no accuracy percentage at all, and its own blog concedes that body-fat scales “often have a margin of error of 3-5% compared to more sophisticated methods like DEXA scans or hydrostatic weighing.” Take the admission at face value, and note the peer-reviewed band above runs wider.

Withingsadvertises “99% reliability” for its Body Scan against DEXA. The figure is real but it is a correlation (r=0.99, from an unpublished internal study of 80 people). And correlation is not agreement. A scale that reads five points high on everyone still correlates almost perfectly with the truth, because correlation measures whether two numbers move together, not whether they match. In a hospital study that measured agreement directly, three consumer scales, a Withings model among them, showed median fat-mass errors of two to four kilograms low against DEXA, while nailing body weight to within a few hundred grams. Weight is measured; composition is estimated. The box quotes the property that flatters the estimate.

Why this matters more on a GLP-1

In pooled DEXA and MRI trials, roughly 25–39% of the weight lost on these medications was lean mass rather than fat. And in arms that added a structured resistance-training programme, the share was far lower. A sustained appetite-driven energy deficit is the whole point of these medications, which makes the fat-versus-lean split worth watching continuously rather than twice a year. And the split is exactly what no single instrument delivers alone. The accurate one is too infrequent to draw a trend; the frequent one is too approximate to trust in absolutes. The lean number also sizes the protein target: a free calculator runs that rule.

So the answer to “DEXA or smart scale” is both, on one chart, which is the chart Bild Health draws. For the hardware decision, our smart scale buyer's guide covers four picks at four budgets, and our guide to keeping muscle on a GLP-1 covers what to do with the numbers once you have them.