Body fat percentage answers the question BMI cannot: what is the weight actually made of.
The catch is measurement. The number your bathroom scale gives you is an estimate produced by passing a small current through your body, and when researchers have checked those estimates against a proper scan, the gap is larger than the decimal place suggests.
Key takeaways
Home bioimpedance underestimated body fat by about 4.6 percentage points in men and 5.1 in women against DXA (Li et al. 2013, PLOS ONE).
The limits of agreement spanned roughly 14 percentage points, so any single reading is soft.
It is far better at tracking change in one person than at stating a level.
Readings swing day to day with hydration, because the method is essentially measuring water.
Why measure body fat at all?
Because total weight hides the thing you usually care about.
BMI cannot distinguish muscle from fat, and misses obesity often enough to matter: in one comparison against DXA scans it classified 26% of adults as obese while the scan found 64%, misclassifying 48% of women (Shah and Braverman 2012). We covered that in BMI: what it's for and where it fails.
Body fat percentage addresses that directly. It also explains the situation that most frustrates people: weight staying flat while the body genuinely changes. In a four-week trial, participants in a substantial deficit doing intense training gained lean mass while losing fat, which scale weight alone would have described misleadingly.
How accurate is a body fat scale?
Directionally useful, numerically soft, and biased in a consistent direction.
A study compared bioimpedance against DXA in adults during a weight-loss programme (Li et al. 2013):
Mean bias | Limits of agreement | |
|---|---|---|
Men, at baseline | 4.6 points underestimated | −2.4 to 11.7 |
Women, at baseline | 5.1 points underestimated | −2.4 to 12.7 |
Men, tracking change | 3.2 points | −4.8 to 11.3 |
Women, tracking change | 3.0 points | −4.8 to 10.7 |
Two things follow. The bias is downward: these devices tend to tell you that you have less body fat than a scan would find. And the limits of agreement span roughly 14 percentage points, which is enormous relative to the range anyone cares about.
So a home reading of 22% is compatible with a true value in the high twenties. Treat the absolute number as an estimate with a wide error bar, not a measurement.
Why does it jump around day to day?
Because bioimpedance is measuring how easily a current passes through you, and that depends mostly on water.
Fat conducts poorly; lean tissue, which is largely water, conducts well. The device infers your composition from resistance, which means anything that shifts your fluid balance shifts your reading:
Hydration. Dehydrated reads as fatter; well hydrated reads as leaner.
Time of day. Fluid redistributes overnight, so morning and evening readings differ systematically.
Food and salt. Both change fluid retention, sometimes for a day or more.
Exercise. Sweat loss and post-exercise fluid shifts both interfere.
Menstrual cycle. Measured with body composition analysis, weight rose about 0.45 kg during menstruation driven almost entirely by extracellular water.
Skin temperature and foot contact. Cold or dry feet weaken the signal on a standing scale.
None of that is fat appearing or disappearing overnight. It is the water term moving, and the device attributing it to composition. The same physics is why daily weight swings by around half a percent of body weight for reasons unrelated to fat.
How should you actually use it?
As a trend from a consistent ritual, which is the only way the error becomes tolerable.
Same time, same conditions. Morning, before eating or drinking, after using the bathroom, before exercise. Consistency matters more here than for almost any other metric.
Same device, always. Different devices use different equations, so switching resets your baseline entirely.
Weekly averages, not daily readings. Given a 14-point spread in agreement, a single reading is close to meaningless while a multi-week direction is informative.
Judge change, not level. The bias is fairly consistent within a person, so a downward trend is real even if the absolute number is wrong.
Pair it with something dumber. A tape measure at the waist has no hydration problem and costs nothing.
What is a healthy body fat percentage?
Wider than most charts imply, and genuinely different by sex and age.
Women carry substantially more essential body fat than men as a matter of physiology, so any single target applied across sexes is wrong. Ranges also shift with age, and the published category boundaries vary by source because they are conventions rather than physiological thresholds.
Given that home measurement is biased low by around five points with a wide spread, comparing your scale's number against a published chart is stacking an imprecise measurement on top of an arbitrary boundary. Your own direction over months is the part you can actually trust.
If you want a fitness measure that is not about body composition at all, VO2 max is worth watching, and pooled analysis found people who were overweight but fit had risk similar to normal-weight fit people.
Frequently asked questions
How accurate are body fat scales?
Not very, in absolute terms. Bioimpedance underestimated body fat by about 4.6 percentage points in men and 5.1 in women compared with DXA, with limits of agreement spanning roughly 14 points (Li et al. 2013). They are considerably better at tracking your own change than at stating a level.
Why does my body fat percentage change overnight?
Because the measurement is driven by how easily current passes through you, which depends mostly on body water. Hydration, food, salt, exercise, time of day and the menstrual cycle all move it without any fat changing.
When should I measure body fat?
In the morning, before eating or drinking, after using the bathroom and before exercise, on the same device every time. The ritual matters more than the moment, because consistency is what makes a noisy measurement readable as a trend.
Is body fat percentage better than BMI?
It measures something more relevant, but it is measured worse at home. BMI is precise and often wrong about what your weight is made of; a body fat scale is directly relevant and imprecise. Using both as personal trends, alongside a waist measurement, is more informative than arguing about which single number wins.
What is a good body fat percentage?
There is no single answer, and it differs substantially by sex and age. Published category boundaries are conventions rather than physiological thresholds, and given that home devices are biased low by roughly five points, comparing your reading to a chart compounds two sources of error. Your own trend is the more trustworthy signal.
Knit shows patterns, not diagnoses. Talk to a clinician about readings that concern you.
