How accurate is the body fat percentage on a smart scale?
Short answer · checked 2026-08-24
Weight is measured; body fat is not. A bathroom scale infers it from one impedance reading and a proprietary equation, and impedance tracks body water: on the standard 73.2% hydration constant, a 0.6 L error in the water estimate moves an 80 kg adult's readout about one percentage point, the same as losing a kilogram of fat. Read the trend at a fixed time of day. Ignore the absolute number.
What we compared
Listed as tested, not ranked
| # | Subject | Type | Official site |
|---|---|---|---|
| 1 | Withings | Device | withings.com |
| 2 | Renpho | Device | renpho.com |
| 3 | Eufy | Device | eufy.com |
| 4 | Garmin Index | Device | garmin.com |
Two numbers, one sensor, and only one of them is a measurement
A body-composition scale contains load cells and an impedance circuit. The load cells weigh you. Everything else on the readout — fat percentage, muscle mass, bone mass, body water, visceral fat, “metabolic age” — is computed from one impedance reading plus your height, age and sex, run through whichever regression equation the manufacturer wrote or licensed. One measurement, six or seven outputs. The outputs are not independent of each other, and none of them is independent of how much water you are carrying.
| What the readout shows | What the hardware did | What the number is |
|---|---|---|
| Body mass | Load cells under the platform | A measurement |
| Body fat % | One impedance reading, foot to foot | A model output |
| Muscle mass | The same reading | Derived from the same output |
| Body water | The same reading | The quantity the model is actually estimating |
| Bone mass | The same reading | Derived, and not what a bone-density scan measures |
| Visceral fat, metabolic age | The same reading, plus age, height, sex | An index, not a unit |
That is not a criticism of any particular product. It is the architecture of the category, and it is why a scale can print 18.4% to a tenth of a point while carrying an uncertainty several times larger than the digit it is showing you. Resolution is not accuracy.
Why the number moves on a day you did not change
Bioimpedance measures resistance to a small alternating current. Fat conducts poorly. Lean tissue, which is mostly water, conducts well. So the quantity the circuit responds to is body water, and the conversion from water to fat-free mass is where the day-to-day noise enters. Two-compartment models assume fat-free mass is 73.2% water — the Pace and Rathbun constant, published in 1945, and still sitting underneath most of these equations.
Work the arithmetic through. If the water estimate is off by ΔW litres and your mass is measured correctly, the fat-free mass estimate moves by ΔW / 0.732, and the displayed fat percentage moves by that divided by your mass:
| Error in the water estimate | Shift in estimated fat-free mass | At 60 kg | At 80 kg | At 100 kg |
|---|---|---|---|---|
| 0.25 L | 0.34 kg | 0.57 pp | 0.43 pp | 0.34 pp |
| 0.50 L | 0.68 kg | 1.14 pp | 0.85 pp | 0.68 pp |
| 1.00 L | 1.37 kg | 2.28 pp | 1.71 pp | 1.37 pp |
| 1.50 L | 2.05 kg | 3.42 pp | 2.56 pp | 2.05 pp |
| 2.00 L | 2.73 kg | 4.55 pp | 3.42 pp | 2.73 pp |
Arithmetic, not a measurement: ΔBF% = (ΔW / 0.732) / mass. pp = percentage points of body fat. Check it yourself; that is the point of showing the expression.
Now the comparison that decides the question. An 80 kg adult at 20.0% body fat who loses one kilogram of fat and nothing else reads 19.0% afterwards — a shift of 1.01 points. From the table, a 0.6 L error in the water estimate is worth 1.02 points. On the display, six hundred millilitres of water and a kilogram of fat are the same event. One of them took a month.
Things that move body water by that much inside a single day: a large drink, a meal, a hot shower, sweat, a night of respiratory and insensible loss, salt, alcohol, the menstrual cycle, and glycogen, which binds roughly three grams of water per gram stored. None of that requires anything unusual to happen to you.
Then there is redistribution, which does not change your total water at all. Foot-to-foot impedance is dominated by the legs. Fluid moves downward across a day spent upright, the legs read as more conductive by evening, and the scale reports a lower fat percentage for the same body. Fixing the time of day is not a tidiness rule. It is the single largest controllable term in the error budget.
What a reference method would have to be
The reason this desk publishes no body-fat accuracy figure is not modesty. It is that we do not own a criterion, and neither does anyone comparing two consumer scales against each other.
| Reference method | What it physically measures | The assumption underneath it |
|---|---|---|
| Hydrostatic weighing | Body density, from mass measured underwater | Fixed densities for fat and lean tissue; residual lung volume must be measured, not guessed |
| Air-displacement plethysmography | Body volume in a sealed chamber | The same fixed-density assumption; thoracic gas volume is estimated |
| Dual-energy X-ray absorptiometry | X-ray attenuation at two energies, region by region | Manufacturer- and software-version-specific; two machines do not necessarily agree on one person |
| Four-compartment model | Density, total body water and bone mineral combined | Fewest assumptions, and the usual criterion. Three instruments, a technician, and a booking |
| Skinfold callipers | Subcutaneous fat thickness at fixed sites | A regression from those sites to whole-body fat; operator-dependent |
| Foot-to-foot bioimpedance | Impedance across the lower body | That you are normally hydrated and that the maker’s equation fits you |
Reviewed 2026-08-10. The table describes methods, not results; no figure in it is ours.
Two consequences follow, and the second one is the useful part.
The first is that “accurate to within X points” is not a property a scale can have on its own. It is a statement about agreement with one reference, on one population, on one day.
The second is that a scale can be biased by several points and still be worth owning, because a constant offset cancels out of a difference. If the equation is wrong for you but wrong in the same direction and by about the same amount every time, then Monday-to-Monday is informative even though no individual Monday is. That is the honest case for these devices, and it is a real one.
The conditions that make the trend usable
A constant offset only stays constant if the conditions do. Everything below holds the measurement path still.
- Same time of day, first thing. After the toilet, before you eat or drink. This controls the largest term.
- Bare feet, dry, on a hard level floor. Damp soles change the impedance path. Carpet changes how the load cells are loaded, which corrupts the one number that was otherwise reliable.
- Same scale, same user profile, same units. Never compare a reading from one brand against a reading from another. The equations differ, so the gap between them is not a change in you.
- Weigh daily, read weekly. Compare the seven-day mean with the previous seven-day mean. A single reading carries the full day-to-day water swing; an average of seven carries much less.
- Ignore weekly moves under about one percentage point. From the arithmetic above, that band is what a normal half-litre hydration swing produces on its own.
- Watch mass and percentage together. If mass is noisy, the percentage will be noisier, because it inherits the mass error and adds the model error on top.
What we will publish on these four, and what we will not
Withings, Renpho, Eufy and Garmin Index are the four brands this desk has queued for the scale bench. Nothing on this page is a measurement of any of them, and none of them is ranked here, because we have not run them yet. What we have declared, in advance, is what the run will and will not produce, under method clause M-06:
- Mass accuracy and linearity against certified calibration masses across the working range, plus repeatability over ten consecutive placements at a fixed load. Published per unit, in grams, with the load at which it was measured.
- Bioimpedance repeatability. The same person, the same scale: five readings inside one session, then morning and evening on the same day across a pre-declared number of days. The figure is the spread, in percentage points. Repeatability we can measure without a criterion; accuracy we cannot.
- No body-fat accuracy figure, for any of them. Not a ranking, not a “closest to DXA”, not a winner. We do not own a reference method for body composition and we are not going to promote a second consumer device into one.
- Prices, read at the point of purchase and dated, under clause M-08.
Declaring that in advance means we cannot quietly acquire a body-fat accuracy claim later because the data looked tidy.
Where this report could be wrong
Not one figure on this page was measured by this desk, and it should be read accordingly. The arithmetic is arithmetic and you can check it in a minute. Everything else here is structural: it describes how these devices work, not how any particular one performs.
Three soft spots. The 73.2% hydration constant is a population average from 1945 and is not exactly right for anybody; multi-frequency devices and proprietary equations do not all lean on it equally hard, and where a maker publishes its equation we will say so and redo the arithmetic with it. Second, devices with hand electrodes measure a different current path and are less exposed to the leg-pooling term; we have tested none of them. Third, the one-percentage-point weekly threshold above comes from the hydration arithmetic at 80 kg, not from a repeatability run. It gets replaced by a measured number when the bench work is done, and if that number is larger we will say so here.
If a manufacturer has a validation against a criterion method that we have missed, send it. Anything that changes a figure here goes in the corrections log with a date on it, not into a silent edit.
Questions we got
Sent to the desk · answered in full
Are smart scale body fat percentages accurate?
Not in the sense the display implies. The percentage is not measured; it is calculated from a single bioimpedance reading, your height, age and sex, and a regression equation the manufacturer chose. That reading follows your body water, and a 0.6 L error in the water estimate is worth about one percentage point of body fat on an 80 kg adult. The number also inherits whatever mismatch exists between you and the population the equation was fitted on, and that part does not average out. What survives is the difference between two readings taken under identical conditions.
Why does my body fat percentage go down during the day?
Because a bathroom scale sends its current up one leg and down the other, so what it mostly measures is the water in your legs. Standing and sitting move fluid downward over the course of a day. More water in the legs reads as lower impedance, lower impedance reads as more lean tissue, and more lean tissue reads as a lower fat percentage — with no change in your body composition at all. This is the single biggest reason to fix the time of day and never compare a morning reading with an evening one.
Which smart scale is the most accurate for body fat?
We are not going to name one, and we would treat any page that does with suspicion unless it states which reference method it compared against and on how many people. Deciding that question requires a criterion measurement this desk does not own — realistically a four-compartment model, which needs three instruments and a technician. Method clause M-06 commits us to publishing bioimpedance repeatability and mass accuracy, and to publishing no body-fat accuracy figure at all until that changes.
Is the weight on a smart scale accurate?
Body mass is the one number on the display that comes off a sensor rather than out of an equation, and it is the number worth trusting on these devices. Two cautions. Resolution is not accuracy: a scale that displays to 0.05 kg is telling you the size of its display step, not its error against a certified mass. And a scale on carpet or an uneven floor loads its cells unevenly, which is a real error and a common one. We will publish mass error against certified calibration masses for the four brands named here when that run finishes.
How often should I weigh myself to see a real change in body fat?
Daily, at the same time and under the same conditions, then read the weekly average rather than any single day. Week-on-week moves under about one percentage point sit inside the band that ordinary hydration swings produce, so they should not be interpreted. Losing four kilograms of fat shifts an 80 kg adult from roughly 20.0% to roughly 15.8%, which is about four points and well clear of the noise — but it takes months, and it is visible in the eight-week trend rather than in this morning's number.
Do the expensive scales with a hand grip work better?
A device that adds hand electrodes measures a current path through the arms and trunk as well as the legs, which removes one specific source of drift: the fluid that pools in your lower body during the day. That is a genuine difference in the measurement, not a marketing tier. It does not remove the hydration dependence, because that sits in the equation rather than in the electrodes, and it does not turn the output into a measured quantity. This desk has not tested one, so treat that as physics rather than as our result.