Body fat percentage: which method to actually trust

Weigh yourself on three different body-fat scales and you will get three different answers. Add a caliper reading and a tape measurement and the spread can easily exceed eight percentage points. This is not because the methods are broken. It is because all of them are estimates, and the marketing rarely says so.

Why bother measuring at all

Bodyweight alone cannot distinguish between losing fat and losing muscle. Two people who both drop four kilograms have had very different outcomes if one lost mostly fat and the other mostly lean tissue.

A body-fat figure also improves the calculators you are likely using. It unlocks lean-mass based energy equations, and it lets protein targets scale off lean mass rather than total weight — which matters a great deal for anyone carrying significant fat, since fat tissue contributes almost nothing to protein requirement.

The realistic options

Tape measurements (Navy method). Free, needs nothing but a tape, and typically lands within about three to four percentage points — the equations were calibrated against hydrostatic weighing at a correlation of roughly 0.90. ¹ Accuracy depends far more on measuring the same anatomical sites each time than on technique subtleties. This is the default we recommend for most people.

BMI-based estimates (Deurenberg). Requires no measurement beyond height and weight, and carries a wider error range — the original paper reports a standard error of about four points, and notes that the formula slightly overestimates body fat in obese subjects, so the error is not evenly distributed across the range. ⁴ Its fundamental limitation is structural: it cannot distinguish muscle from fat, so it systematically overestimates for muscular people and underestimates for the very sedentary.

Skinfold calipers (Jackson-Pollock). The source equations report a standard error around three and a half points for men ² and slightly wider for women ³. Accurate in trained hands and unreliable in untrained ones. The equations themselves are sound; the variance comes from whether you pinch the same spot, at the same depth, with the same pressure, every time.

Bioelectrical impedance scales. Convenient and heavily influenced by hydration status. Take one reading in the morning and another after a litre of water and you can shift the result by several points without changing anything real. Useful for trend, poor for absolute value.

DEXA. The practical reference standard, and still not perfect — readings vary between machines and even between scans on the same machine. It costs money and requires a booking, which is why almost nobody uses it for routine tracking.

Repeatability beats accuracy

Here is the point most articles on this topic miss.

Suppose your true body fat is 18 percent and your tape measurement consistently reads 21. That method is wrong by three points — and yet it is entirely fit for purpose, because when you actually drop to 15 percent, it will read 18. The error is a constant offset, so the change it reports is correct.

Now suppose you use a different method each month. One reads 21, the next 16, the next 19. You have learnt nothing about whether you are making progress, despite each individual reading being no less accurate than the tape.

Pick one method. Measure under the same conditions — same time of day, same hydration state, same sites. Compare only to your own previous readings from that same method, and ignore comparisons to anyone else’s number.

Using the number sensibly

Do not chase precision you cannot have. If your method carries a ±3.5 point error, treating a shift from 18.2 to 17.9 as meaningful is reading noise. Look for changes larger than the error range, over at least a month.

If you are guessing wildly, leave it blank. Calculators that accept a body-fat input will happily use a bad one. An unreasonable figure produces worse targets than no figure at all, because the tool will confidently scale protein and energy from it.

Consider waist-to-height ratio alongside it. It requires one measurement, has no estimation error to speak of, and predicts cardiometabolic risk better than BMI. A ratio above 0.5 is the commonly cited threshold. ⁵ It tells you something different from body fat percentage, and it is harder to get wrong.

The bottom line

There is no consumer method that will tell you your body fat percentage to within a point, and the ones that claim to are overselling. Accept the error range, pick the method you will actually repeat — for most people that is a tape measure — and standardise how you use it.

Then largely ignore the absolute number and watch the direction it moves over months. That direction is the only part of this that reliably means anything.