Body fat testing: how wrong are scales, calipers and DEXA?

Your bathroom scale says 18.4% body fat in the morning and 21.1% the same evening. A DEXA scan puts you at 24%. Your coach, with a set of calipers, gets 16%. The useful question isn't which number is correct — it's how much any of them actually knows.

The honest answer is: less than most people assume, but not so little that measuring is pointless. What matters is knowing what each device is really reading, and what you then do with the number.

What we know reliably

None of these tools measures fat. The evidence here is as solid as it gets, because this isn't a research question — it's how the machines work. A scale measures electrical resistance. A caliper measures a fold of skin in millimetres. DEXA measures how much X-ray energy at two different levels each part of your body absorbs. All three feed that signal into an equation built on a sample of people, and hand back an estimate. If you don't resemble that sample, the estimate drifts.

The accuracy ranking is settled. Best to worst: laboratory multi-compartment models that measure water, mineral, protein and fat separately; then DEXA and underwater weighing; then calipers in trained hands; then tape measurements; and last, a home bioimpedance scale. That order shows up consistently enough that you can treat it as fixed.

Home scales are hostage to your hydration. Current passes more easily through water-rich tissue, so anything that shifts your fluid balance shifts the reading: a glass of water, a salty dinner, a training session, a sauna, last night's beer, where you are in your menstrual cycle. Swings of two to four percentage points within a single day are ordinary. Against laboratory methods, the error for an individual is often around four to five percentage points, and larger for some people. On top of that, a scale you stand on reads mostly your lower body while a handheld device reads mostly your upper body, so two devices in the same house will disagree by design.

With calipers, the error lives in the tester. A practised measurer who always pinches the same site, on the same side, at the same time of day gets impressively stable numbers — repeat measurements usually land within about a millimetre per site. Two different testers on the same person can diverge far more. Converting millimetres into a body fat percentage adds another layer of error, because the equations assume a fixed relationship between fat under the skin and fat inside you, which varies between individuals.

DEXA is a good reference, not the truth. Two machines from different manufacturers — or even two software versions on the same machine — can return numbers several percentage points apart for the same person on the same day. Two scans back to back are rarely identical either. Where DEXA genuinely excels is tracking change in one person, on one machine, under matched conditions.

Probable, but not settled

How far off your particular device is for you specifically is unknown. Published figures are group averages, and the spread around them is wide. One person lands within a percentage point of the reference; the next is out by seven.

Standardising conditions almost certainly cuts the noise on bioimpedance: first thing in the morning, fasted, before drinking, before training, no alcohol the night before. The direction is clear; exactly how much each condition contributes on consumer scales is poorly quantified.

Glycogen holds water, so after a few high-carbohydrate days DEXA typically shows more lean mass and sometimes a lower fat percentage without a gram of real change. The mechanism is convincing, the size of the effect in one person isn't. The same caution applies to measuring straight after a fasted session or a long fast.

The segment readouts your scale offers — leg fat, trunk fat, a "visceral fat" index — are barely studied on consumer devices. Treat them as a curiosity, not data.

What gets misreported

"DEXA is the gold standard." It isn't, though it's sold that way. The gold standard is a laboratory multi-compartment model, and DEXA typically departs from it by two to four percentage points — not always in the same direction.

"I dropped 3% body fat in a fortnight." On a realistic deficit, fat loss runs roughly half a kilo to a kilo per week. Three percentage points in two weeks for an 80 kg person would mean about 2.5 kg of pure fat. It's almost always water plus measurement noise.

Mixing sources into one trend line. A caliper reading from last year and a DEXA scan from today don't sit on the same graph. They're two different scales, and the gap between them means nothing.

Advertised accuracy of ±1%. Claims like that on home scales have no independent backing. The same reflex that helps you read a supplement study without falling for the headline applies here: a number that sounds too precise is usually marketing.

"Body fat percentage tells me how healthy I am." It tells part of the story. Waist circumference, blood pressure, blood work, sleep and daily activity carry at least as much information, often more.

What this means for your training and diet

The working rule: pick one method, standardise it, follow the trend, and ignore the absolute figure.

• Bioimpedance scale: same time of day, morning, after the loo, before food and drink, no training in the previous 12 hours. Measure three times a week and read the average, never a single number.
• Calipers: same tester, same sites, every time. Record the sum of the folds in millimetres and track that sum. Converting to a percentage adds error without adding information.
• DEXA: no more than every three to four months, same machine, same conditions — fasted, with the same routine the day before. Less often, but cleaner.
• The free markers: body weight as a seven-day average, waist measured at the navel, photographs every four to six weeks, and your numbers on the main lifts.

Never change a plan on the strength of one measurement. Wait until at least two markers move the same way for three to four weeks. If weight, waist and strength are all flat, that's when you touch calories.

And don't blame the measurement when the slowdown has another cause. As a diet drags on, your total daily energy expenditure genuinely falls, which has nothing to do with how precise your calipers are. What protects muscle in a deficit is serious strength training and enough protein — usually 1.6 to 2.2 grams per kilogram of body weight per day.

When to see a doctor

• You've lost more than 5% of your body weight over six to twelve months without trying.
• Swelling in the legs or sudden fluid retention. At that point the accuracy of your body composition reading is the least of it — that's a medical signal.
• Periods stop, fatigue won't lift, injuries stack up, or strength drops and rest doesn't fix it. Before you blame the scale, check the signs of overtraining.
• You have a pacemaker or another implanted electronic device — manufacturers of bioimpedance scales normally list this as a reason not to use them.
• Pregnancy: the radiation dose from DEXA is very small, but body composition scanning is postponed because there's no medical reason for it.

In short

• No device measures fat; all of them estimate it from an indirect signal plus an equation.
• Home scales are the weakest for an absolute number and the most sensitive to water; four to five percentage points of error is normal.
• Calipers work well in one trained pair of hands — track millimetres, not percentages.
• DEXA is the best widely available option but it isn't truth: different machines give different numbers for the same person.
• Every method is far better at tracking change than at naming an absolute figure.
• Standardise conditions, follow the trend over several weeks, and only act when more than one marker agrees.