Understanding body-fat estimates and their error bars
Every body-fat method is an estimate with an error bar of at least 1–3%, and for bioimpedance scales it is 4–8% and depends on hydration. Which means a scale reporting 18% could reasonably mean anything from 12% to 24%. Use the same method repeatedly to track direction; do not trust any single absolute number.
There is no direct measurement
The first thing to understand is that nothing available to you measures body fat. Every method measures something else — electrical resistance, skinfold thickness, X-ray attenuation, body volume — and infers fat from it using assumptions about tissue density and hydration.
Those assumptions are population averages. You are an individual. The gap between the two is the error bar.
| Method | Typical error | Cost | Verdict |
|---|---|---|---|
| DEXA scan | ±1–3% | High, clinic visit | The practical reference standard. Still not exact. |
| Hydrostatic weighing | ±2–3% | High, rare | Accurate, impractical. |
| BodPod | ±2–4% | High | Accurate-ish, affected by breathing and clothing. |
| Skinfold callipers, trained tester | ±3–5% | Low | Good for tracking change if the same person measures. |
| Skinfold callipers, self-administered | ±5–8% | Low | Poor absolute accuracy, usable for trend. |
| Tape-measure formulas (Navy method) | ±4–6% | Free | Cheap and surprisingly serviceable for tracking. |
| Bioimpedance scale | ±4–8%, hydration-dependent | Low | Poor. Use the weight, ignore the body-fat number. |
| Handheld bioimpedance | ±5–10% | Low | Worse than scales — it measures mainly the upper body. |
| Visual estimation from photos | ±5% | Free | Better than people expect, with good reference images. |
Read the bioimpedance row carefully, because it is what most people actually use. A ±4–8% error on a reading of 18% means the true value plausibly sits between 10% and 26%. That range spans "lean" to "average" — the number is not telling you what you think it is telling you.
Why bioimpedance is so unreliable
Bioimpedance sends a small current through the body and measures resistance. Fat conducts poorly, lean tissue conducts well — so more resistance implies more fat. The problem is that conduction depends overwhelmingly on water, and your hydration state changes constantly.
Things that change a bioimpedance reading without changing your body fat at all:
- How much you have drunk in the last few hours
- Whether you have eaten
- Whether you have trained today
- Whether you have showered (warm skin, altered blood flow)
- Time of day — readings drift substantially from morning to evening
- Alcohol the night before
- Menstrual cycle phase
- Carbohydrate intake, through glycogen and its associated water
- Whether your feet are dry, for foot-to-foot scales
Handheld devices are worse still, because current takes the path of least resistance through the upper body and largely bypasses the legs — so they systematically misestimate anyone whose fat distribution is not average.
If you own a bioimpedance scale, it is still useful. The weight it reports is accurate. Use that, and ignore the body-fat percentage — see weekly averages versus daily weigh-ins.
Even DEXA is not truth
DEXA is treated as the gold standard and is frequently reported as though it were exact. It is the best practical option and it still carries a ±1–3% error, plus some specific quirks:
- Hydration affects it too, though less than bioimpedance.
- Different machines and software versions disagree. Comparing a scan from one clinic with one from another is unreliable.
- Positioning matters. Small differences in limb placement change the result.
- It cannot distinguish some tissue types as cleanly as its precision implies.
- Reported to one decimal place, which conveys a false sense of precision.
Two DEXA scans on the same machine, same conditions, weeks apart, are a reasonable way to measure change. A single DEXA scan is a reasonable estimate with an error bar, not a fact about you.
What to do instead
The practical position: stop trying to know your body-fat percentage, and start tracking whether it is going the right way. Those are different problems, and the second one is both more useful and much easier.
- Waist circumference. Measures a physical dimension directly, no assumptions involved, error of about ±0.5 cm with good technique — see how to measure your waist consistently.
- Photos. Show fat distribution and change more honestly than a percentage does. Four-weekly, standardised.
- Weight trend. Accurate to the gram, as a weekly average.
- Your logbook. Tells you whether you are keeping muscle while losing weight — see tracking strength progress over months.
- The same body-fat method, repeatedly, if you want a number. A bioimpedance scale used every morning under identical conditions tracks direction adequately even though the absolute value is wrong.
Reading the numbers you are given
If you do get a body-fat estimate, some interpretation guidance:
- Report it as a range. "Roughly 15–20%" is honest. "17.4%" is not.
- Only compare within a method. A DEXA result and a scale result are not comparable numbers.
- Only compare within a device. Two different scale brands will disagree by several points.
- Ignore changes under 2–3 points from any cheap method. That is inside the error.
- Standardise the conditions if you are tracking with bioimpedance: same time, same hydration state, same day of the week.
- Essential fat differs by sex. Roughly 3–5% for men and 8–12% for women, which is why the same percentage means very different things.
Note also that reference categories vary between sources. Broadly, athletic ranges are around 6–13% for men and 14–20% for women, with fitness ranges above that — but these bands are approximate and the measurement error means you cannot place yourself precisely within them anyway.
Common questions
How accurate are body-fat scales?
±4–8%, heavily hydration-dependent. Use the weight, ignore the percentage.
Is DEXA worth paying for?
Occasionally, as a reasonable estimate. Two scans months apart on the same machine are more informative than one.
Why does my scale give different readings within an hour?
Hydration, temperature and blood flow. It is measuring water more than fat.
Do calliper measurements work?
Reasonably, for tracking change, if the same trained person measures every time.
What about tape-measure formulas?
Surprisingly serviceable for tracking, and free. The Navy method uses neck, waist and (for women) hips.
Should I care about my body-fat percentage at all?
Mostly no. Waist, photos and your logbook answer the questions you actually have.
Why did my body fat go up when I lost weight?
Almost certainly measurement error or hydration, not a real change.
The practical version
- No available method measures body fat directly; all infer it from something else.
- Bioimpedance scales carry ±4–8% error and depend heavily on hydration.
- DEXA is the best practical option at ±1–3%, and still not exact.
- Ignore changes under 2–3 percentage points from any cheap method.
- Only compare within a single method and a single device.
- Waist, photos, weight trend and your logbook answer the real questions better.
Key takeaways
- Nothing available measures body fat directly — every method infers it with assumptions.
- Bioimpedance scales carry ±4-8% error, so an 18% reading could mean 10% to 26%.
- Bioimpedance tracks water more than fat, so hydration, food and training all shift it.
- DEXA is the best practical option at ±1-3% and still not a fact about you.
- Ignore changes under 2-3 percentage points from any inexpensive method.
- Waist circumference measures a physical dimension directly, with error of about ±0.5 cm.
Related reading
General information, not medical or individualised advice. Speak to a doctor before starting a new programme, especially if you have a medical condition, are pregnant, or are returning from injury. If you train with a coach, their guidance takes precedence.
Track it, or none of this applies
MyFitsy pre-fills the weight you used last time, so progression is a decision instead of a guess.
