Topline
Calipers, tape, smart scales and BMI-based equations all claim to measure body fat. Here is how each one works, how wrong it typically is, and which to trust.
Every method available to you at home is an estimate produced by an equation, not a measurement of fat. None of them observes adipose tissue. They observe something correlated with it, then apply a regression fitted to a reference population, and the gap between that population and you is where the error lives.
That framing matters more than the ranking, because it changes what you should ask of the result. The right question is not which method gives the true number. It is which method gives a number that is wrong in a consistent direction, because a consistently wrong number still tracks change accurately, and change is what almost everyone actually wants to know.
What accuracy means when there is no perfect reference
Even the laboratory standards disagree with each other. Underwater weighing and air displacement plethysmography measure whole-body density, then convert it to fat percentage using the two-compartment model Siri published in 1961. That conversion treats fat-free mass as having a constant density, when its water and mineral content in fact varies with age, hydration, ethnicity and training history.
DEXA takes a different route, separating tissue by X-ray attenuation, and is the usual practical reference. It is not a gold standard either: results differ measurably between manufacturers and between software versions on the same machine, so a DEXA figure of 22% carries a percentage point or two of its own uncertainty before your body enters the picture.
So when a method is described as carrying three to four percentage points of error, that error is measured against an imperfect reference. Read published accuracy figures as the spread to expect around a single reading, not as a distance from a knowable truth.
The ranking, in short
| Method | Typical error vs DEXA | Cost | Main weakness |
|---|---|---|---|
| Skinfold calipers, 3 or 7 site | ~3 to 4 points, better with a skilled tester | Low | Technique dependent; needs practice |
| Tape circumference, Navy method | ~3 to 4 points | Very low | Ignores limbs and muscularity |
| Bioelectrical impedance, scales | ~4 to 8 points | Low to moderate | Hydration sensitive; poor at extremes |
| BMI-based equations | ~4 to 5 points | Free | Cannot distinguish two bodies of equal BMI |
The differences between the top three are smaller than the marketing around them suggests, and smaller than the difference between a careful protocol and a sloppy one within any single method. A well-executed tape measurement beats a badly taken caliper reading comfortably.
Skinfold calipers: best at home, hardest to do well
Calipers pinch a fold of skin and subcutaneous fat at defined sites, on the reasoning that subcutaneous thickness tracks total fatness. The measurements feed a regression that predicts body density, which is then converted to fat percentage.
Two equation families dominate. Jackson and Pollock (1978) published sex-specific equations using three or seven sites, and their three-site version is what most consumer calipers ship with. Durnin and Womersley, in Br J Nutr 32:77-97 (1974), derived age-banded equations from four sites on 481 men and women aged 16 to 72, and their approach is still common in clinical and research settings.
Done properly, calipers are the most accurate option at home. The difficulty is in "properly". Sites must be located precisely, folds lifted the same way each time, and the jaws read at a consistent interval after closing, because the tissue keeps compressing. Inter-observer variation dominates the error, and self-measurement is worse still: you cannot reach your own subscapular or mid-axillary sites reliably, and the sites you can reach are the ones your posture distorts.
If you use calipers, take three readings at each site and use the median rather than the mean, have the same person measure every time, and accept that the first two months of readings are mostly you learning the technique rather than your body changing.
Tape measurements: the Navy method
The circumference approach infers fat distribution from where your body is wide. Hodgdon and Beckett developed the version in widest use in Naval Health Research Center Report 84-29 (1984), fitting the logarithm of waist minus neck against the logarithm of height for men, and adding hip circumference for women.
A tape is cheap and the landmarks are easier to find than caliper sites. The limitation is structural: the equation sees your torso and nothing else. Two men with identical waist, neck and height return the same estimate however much muscle they carry on their arms and legs, which is why the method reads athletic bodies as fatter than they are. It was also fitted on military personnel, younger and less varied in build than the general public.
The sensitivity to tape placement is worth knowing precisely. For a man of 178 cm with a 39 cm neck, our body fat calculator returns 14.9% at an 84 cm waist, 16.4% at 86 cm and 18.0% at 88 cm. That is roughly three quarters of a percentage point per centimetre, so measuring at your navel one week and at your natural waist the next introduces more variation than a month of genuine progress. Neck placement moves it the other way: holding waist at 86 cm, a 38 cm neck reads 17.2% and a 40 cm neck reads 15.7%.
For a woman of 165 cm with a 32 cm neck and 100 cm hips, the same tool returns 29.4% at a 74 cm waist and 31.4% at 78 cm, so the female equation is a little less sensitive per centimetre but adds a second circumference that can drift.
Taking the measurements consistently
Protocol matters more than method. Measure first thing in the morning, after using the bathroom, before eating or drinking, and before any training. Stand relaxed with your arms at your sides and breathe normally rather than holding your breath or bracing your abdomen.
For the waist, men should measure at the narrowest point, usually just above the navel, and women at the natural waist, the narrowest point between ribs and hips. Keep the tape horizontal all the way round, snug enough to sit against the skin without compressing it, and take the reading at the end of a normal exhale. For the neck, measure just below the larynx with the tape sloping slightly downward at the front. For hips, use the widest point of the buttocks.
Write down the landmark you used, not just the number. Six months later you will not remember, and that record is what makes the trend interpretable.
Two comparisons make the record more useful. A waist-to-height ratio needs no equation and no reference population, which is why waist-to-height ratio explained treats it as the most robust thing a tape produces. And lean mass in kilograms, from body weight minus estimated fat mass, is often the more actionable figure during a deficit. What lean body mass is covers why, and it feeds the Katch-McArdle option in our TDEE calculator.
Bioelectrical impedance: convenient, and the most variable
Impedance devices pass a small alternating current through the body and measure resistance. Lean tissue holds most of the body's water and electrolytes and conducts well; fat conducts poorly. The device measures impedance, then applies a proprietary equation involving height, weight, age and sex to predict fat-free mass.
Two things limit it. Hydration changes conductance directly, so the reading responds to how much you drank, recent training, sodium, alcohol and cycle phase; the same body can read several points apart on one day. And geometry works against it: foot-to-foot scales pass current up one leg and down the other, largely missing the torso where most variation in fat storage happens. Hand-to-hand devices have the mirror problem. Eight-electrode devices using both hands and feet perform better.
The equations are proprietary, so you cannot see what population yours was fitted to, and performance is typically worst at the extremes, in very lean and very heavy people, who are the groups most likely to be tracking closely.
None of this makes impedance scales useless. Weighed daily under identical conditions, the trend they produce is informative even when the absolute number is several points off. Treat the daily figure as noise and the thirty-day average as the signal.
BMI-based equations: useful, but not independent
Deurenberg et al., in Br J Nutr 65:105-114 (1991), derived an equation predicting body fat percentage from BMI, age and sex, validated against densitometry in a sample of over 1,200 adults. It requires only a scale and a height measurement.
Understand what it can and cannot do. Because BMI is the only body input, the equation returns the same answer for any two people sharing a BMI, an age and a sex. Two men of 178 cm and 82 kg, one with a 78 cm waist and one with a 97 cm waist, both receive 22.9% at age 35 even though the Navy equation separates them by more than sixteen points. A BMI-derived estimate cannot resolve the disagreement between BMI and body composition, because it is BMI with a demographic correction applied.
The age term is also large enough to notice. Holding BMI at 24 for a man, the equation moves from 18.3% at age 25 to 22.9% at 45 and 27.6% at 65, on the reasonable population assumption that lean mass declines with age. For an individual who has resistance trained through those decades, that assumption does not hold, and the equation will overstate fat accordingly.
Use it as a sanity check against another method rather than as a primary estimate. Where the two disagree sharply, the disagreement itself is the useful information: it usually means your build differs from the population average in a specific, identifiable direction.
Building a protocol that actually tells you something
Pick one method, ideally two, and hold everything else constant. Same time of day, same conditions, same landmarks, same person measuring. Then stop reading individual numbers.
Measure every two weeks rather than daily. Body fat does not change fast enough to be visible over shorter intervals, and at a sustainable rate of fat loss you might expect something in the region of half a percentage point a month, which is comfortably inside the error of every method above. What you can detect over three months is a trend, and a trend is what warrants a decision.
Record raw inputs alongside the output. Waist, neck, hip and weight in centimetres and kilograms are facts about your body; the fat percentage is an inference from them. If you later prefer a different equation, or want to compare against a scan, you can recompute from raw numbers but not from a percentage.
Watch for one artefact. If your estimate improves substantially in the first fortnight, that is usually the method settling rather than your body changing, and it happens most often with impedance scales after a diet change alters glycogen and water. Discard the first two readings of any new protocol.
What none of these methods can tell you
No home method distinguishes visceral fat from subcutaneous fat, and that distinction carries independent risk. Fat around the abdominal organs is metabolically active in a way that fat under the skin is not, and it tracks cardiometabolic outcomes more closely than total fat mass. Only imaging separates the two. Waist circumference is the practical proxy, and a reasonable one.
Nor does any home method have a threshold tied to hard outcomes. The body fat categories our body fat calculator reports, and every set you will find elsewhere, come from fitness and clinical reference conventions rather than the pooled mortality data behind the BMI bands. A reading of 26% for a man sits in a category described as above average; that is a convention, not a risk estimate from cohort studies. Crossing a boundary is not an event.
And none of them is a diagnostic instrument. If a reading concerns you, or composition is changing without an explanation you can identify, that belongs with a clinician who can see markers no tape reaches. Measuring at home gives you a trend to bring to that conversation rather than a substitute for it.