The numbers a smart scale prints: which are measured, which are calculated, and which are the manufacturer's own definition?

Updated: 2026-08-31

For the body composition figures, a smart scale actually measures only two things: your weight and an electrical impedance. (⚠️ Individual models sense other physiological signals as well — one manufacturer's page opened for this article states that a particular model 「has the ability to measure the resting heart rate」 — but those are not among the body composition figures derived from impedance.) Every other body composition number — body fat percentage, muscle mass, bone mass, body water, basal metabolic rate, metabolic age — is calculated by an equation from that impedance plus the age, sex and height you enter. The Consumer Council put it plainly in 2017: "different smart scales may use different equations, so the body composition figures they produce may differ." Manufacturers do not hide this: one manufacturer's own page describes muscle mass as 「predicted」 and bone mass as 「estimated weight of bone mineral」, while metabolic age 「compares your BMR with the average of your age group」 — a comparison, not a measurement. How accurate? A study comparing 15 BIA devices (14 of them consumer-grade) against a laboratory four-compartment model found repeat measurements very stable (precision error 0.0% to 0.49%), but constant error against the reference method ranging from −3.5 to +11.7 percentage points. ⚠️ The authors themselves qualify that stability: five of the 15 were foot-to-foot devices whose least significant change was only 0% to 0.11%, which the authors call 「suspiciously high」 reliability, suggesting these models may include an algorithm preventing differences beyond a certain size on repeat measurement within a short period; the other 10 ranged from 0.38% to 1.36%, which the authors call "reasonable data". The Population Health Survey sets no threshold for the impedance-derived body-composition readings discussed here. Across all 81 pages of that 2023 report, the terms for body fat, muscle mass, bone density, basal metabolic rate and metabolic age do not appear once; the indicators it uses are BMI, waist circumference and waist-to-hip ratio — and of those three, a scale can only compute BMI (and only if you enter your height); waist circumference and waist-to-hip ratio both need a tape measure. ⚠️ Safety note: the Consumer Council states that people with implanted electronic medical devices such as a pacemaker or respirator should not use a smart scale to measure body composition. This article explains where the numbers come from. It rates no brand or model and offers no medical advice.

---

#What does the scale actually measure?

For body composition: two things — your weight and an impedance value. Everything else is calculated. (Some models have other sensing functions, such as resting heart rate, but those are not composition figures derived from impedance.)

From the Consumer Council's feature of June 2017:

"most smart scales now use bioelectrical impedance analysis. Because different tissues in the body impede electric current differently, a smart scale can use the resistance data together with preset equations to measure or calculate body fat percentage, water content, muscle mass, bone mass, basal metabolic rate, body age and so on. Different smart scales may use different equations, so the body composition figures they produce may differ."

Two points in that passage:

One: the principle is impedance. Current passes through the body and different tissues resist it differently. Water and muscle conduct well, fat conducts poorly — what the scale directly measures is the electrical value, impedance; body water itself is derived from impedance, and fat and the other components are derived after that. Two: the equation is the manufacturer's own. The same person standing on two scales of different makes can get different body fat percentages, because the two equations differ.

Who most needs to know this: anyone who sees two scales disagree and assumes one is broken — both may be working, and simply using different equations.

---

#Reading by reading: which is measured, which is calculated

Everything in the "what the manufacturer says" column below is quoted from one manufacturer's own explanatory page. ⚠️ This article opened only one manufacturer's page, so the following represents that manufacturer's definitions of its own readings, not an industry-wide standard.

⚠️ "Body water" deserves particular attention, because it explains why all the other numbers drift. The manufacturer writes: "The body's water content is constantly changing. Water is lost through urine, sweat and breathing, but your hydration level can also vary depending on, for example, alcohol consumption, the flu or menstruation." The same page adds: "Please note that your body water measurement should be used as a guideline…"

What the scale directly measures is impedance; body water is derived from it. So when hydration changes, impedance changes, and the numbers derived from impedance may change with it. ⚠️ But this cannot be put as "they all change together": how much and in which direction depends on the model, the equation, the measurement frequency, the measurement protocol and the physiological state at the time; BMI does not go through impedance at all (only height and weight); and none of the sources opened for this article states that every output must move in step. The manufacturer's own wording is only that body water is "constantly changing" and that the figure is "a guideline".

---

#How accurate? "Precise" and "accurate" are two different things

The point to hold on to: repeat measurements on one machine are very stable, but the machine can be far from the laboratory method.

The study's design and results (Br J Nutr; the journal prints an issue date of 14 September 2023):

"Seventy-three healthy participants aged 19–50 years were assessed by each of fifteen BIA devices… compared with a 4C model… The sample of devices included fourteen consumer-grade and one research-grade model…"

(In plain terms: 73 healthy participants aged 19 to 50 were assessed on 15 BIA devices against the laboratory four-compartment model; 14 were consumer-grade and 1 research-grade.)

The results:

"BIA devices demonstrated high reliability, with precision error ranging from 0·0 to 0·49 %. Cross-sectional validity varied, with constant error relative to the 4C model ranging from −3·5 (sd 4·1) % to 11·7 (sd 4·7) %, standard error of the estimate values of 3·1–7·5 % and Lin's concordance correlation coefficients (CCC) of 0·48–0·94."

In short:

• Precision (repeat measurement on one machine): error of 0.0% to 0.49%. ⚠️ But the authors note that 5 of the 15 foot-to-foot devices had a least significant change of only 0% to 0.11%, "suspiciously high" reliability that may reflect an algorithm suppressing differences on repeat measurement within a short period; the other 10 ranged from 0.38% to 1.36%, which the authors treat as reasonable data. In other words, for 1 model in every three "very stable" may not mean genuinely well measured.
• Accuracy (against the laboratory four-compartment model): constant error from −3.5 to +11.7 percentage points.

⚠️ The two are not contradictory. A ruler can give the same reading every time (precise) while the whole ruler is skewed (inaccurate) — a scale can report a very consistent number daily while being systematically offset from the reference method. ⚠️ This article derives no individual "true value" range from the −3.5 to +11.7 span: constant error is each machine's mean bias, not an individual error bound; the study also reports each model's standard error of the estimate, proportional bias (statistically significant in some of the 15) and limits of agreement, and interpretation at the individual level needs that model's own figures.

The same study is more optimistic about tracking change:

"For longitudinal validity, constant error ranged from −0·4 (sd 2·1) % to 1·3 (sd 2·7) %…"

(For the 37 people remeasured after 12 to 16 weeks, constant error was −0.4 to 1.3 percentage points, standard error of the estimate 1.7% to 2.6%, Lin's CCC 0.37 to 0.78. ⚠️ Constant error is a mean bias; a narrow range does not mean good individual agreement.)

The authors' own conclusion, with their own qualification:

"While performance varied widely across the sample investigated, select models of BIA devices (particularly octapolar and select foot-to-foot devices) may hold potential utility for the tracking of body composition over time…"

The practical conclusion: if your scale is among the better performers in the study, watching the trend is more meaningful than watching a single day's absolute figure. ⚠️ But the study says select models — across the 15, longitudinal agreement (Lin's CCC) ran from 0.37 to 0.78, meaning about half performed only moderately at tracking change. Standardising measurement conditions cannot by itself make an unvalidated scale track reliably.

A second study shows the same thing. A study comparing a standing bioelectrical impedance analyser with dual-energy X-ray absorptiometry (DXA) is titled "High precision but systematic offset". The authors found the device "underpredicted %BF (bias −3.4 and MAE 3.9%)", and noted that "A 3% correction factor applied to the InBody device results provided values very close to the DXA measurements". ⚠️ But the scope must be stated: that study compared a standing research/clinical-grade analyser, not a household scale, and its 3% correction factor cannot be transferred to the scale in your bathroom.

Who most needs to know this: anyone who weighs themselves daily and is delighted when body fat drops 1% — 1% sits within the error range of these studies.

---

#Bone mass is not bone density

The "bone mass" your scale gives is a weight estimate; osteoporosis is diagnosed on bone density, which is a different thing.

The manufacturer's own page is clear that bone mass is "The estimated weight of bone mineral in your body", and how it is produced:

"The body analysis scale calculates your bone mass within seconds. This is done using a statistical calculation based on existing studies. These studies have shown that there is a close correlation between the weight of your bones and your lean mass."

So bone mass is a statistical estimate. ⚠️ The manufacturer discloses only that it is produced by a statistical calculation based on existing studies showing a close correlation between bone weight and lean mass. It does not publish the equation, and does not say lean mass is the sole input; this article draws no such inference.

How accurate is that kind of estimate? A study comparing bioelectrical impedance with DXA for bone mineral content (302 healthy Korean participants) found:

"Age-specific regression models significantly enhanced prediction accuracy… In contrast, existing BIA equations exhibited substantial bias (mean difference up to 0.46 kg, p < 0.001)."

(In plain terms: existing BIA equations showed bias of up to 0.46 kg on average.)

The same paper states its own limitations and conclusion: "limitations include reliance on a single BIA device, lack of external validation, and an unbalanced age distribution… While DXA remains the gold standard…"

⚠️ One point to note: that study concerns bone mineral content (a weight), which is what the scale reports; osteoporosis is diagnosed on bone mineral density, which is another thing. A scale cannot tell you about osteoporosis.

---

#Which indicators do officials use? Not the ones a scale prints

Hong Kong's Population Health Survey uses BMI, waist circumference and waist-to-hip ratio; it sets no threshold for the body fat percentage, muscle mass, bone mass, basal metabolic rate or metabolic age a scale prints.

⚠️ This absence was checked page by page against that one report, not against every official document in Hong Kong. Across all 81 pages of the Report of Population Health Survey 2020-22 (Part II) (2023), the following terms appear this many times: 「體脂」 0, 「脂肪百分」 0, 「肌肉量」 0, 「骨質密度」 0, 「骨密度」 0, 「基礎代謝」 0, 「代謝年齡」 0, 「生物電阻」 0. 「內臟脂肪」 (visceral fat) appears twice, both times in the passage explaining why waist circumference is used, with no numeric threshold.

What officials actually use:

⚠️ The same population, two classifications, very different answers. From the report: under the Asian adult classification, the age-standardised prevalence of overweight and obesity among people aged 18–84 was 51.3%; under the WHO classification (overweight ≥25.0, obesity ≥30.0), the same population gives 30.8%.

This bears directly on your scale: the manufacturer's page opened for this article prints the BMI thresholds "A number between 18.5 and 25 indicates a healthy weight. A number above 25 is overweight and a number above 30 is obese" — that is WHO's general classification, not the Asian classification the Hong Kong report uses. So: a scale calling your BMI "normal" may already be "overweight" on Hong Kong's thresholds.

What about visceral fat? The report states that WHO recommends waist circumference and waist-to-hip ratio "because these indicators better reflect the mass of visceral adipose tissue" — so the proxy indicators officials use for visceral fat are waist circumference and waist-to-hip ratio, not the level your scale gives. Both need a tape measure; the scale cannot measure them.

One more figure worth remembering: the report found that 65.1% of people classified as overweight considered themselves underweight or about right; among those classified as class I obese, 27.6% did so, and among class II obese, 3.2%. ⚠️ Each percentage has its own denominator (the overweight group, the class I obese group, the class II obese group); none is a share of the Hong Kong population.

---

#How do you get something useful out of it?

The Consumer Council's June 2017 feature gives a specific method, reproduced here.

"place the scale on a level, hard floor and do not stand on its edge. For stable readings, use it at the same time and in the same environment each day, and wait about 3 hours after eating, vigorous exercise or getting up in the morning."

⚠️ Safety note, not to be skipped: the same page states "people with implanted electronic medical devices such as a pacemaker or respirator should not use a smart scale to measure body composition, so that the low-frequency current it emits does not affect the device's operation."

On privacy: the Council also cautions that once a smart scale is linked to a phone app, "users should note the permissions the app requests, and cannot know when it accesses that data or how it is used".

Together with the study's conclusion — in the authors' words, "While performance varied widely across the sample investigated, select models of BIA devices (particularly octapolar and select foot-to-foot devices) may hold potential utility for the tracking of body composition over time". ⚠️ They say "select models", not "this class of device". For tracking change, Lin's CCC ran only 0.37 to 0.78, and some models showed statistically significant proportional bias. So the sensible use is — the same scale, the same time of day, the same state, watching the trend over several weeks; not comparing a single day's absolute figure with another person or another scale. ⚠️ And to be clear: standardising conditions can only reduce avoidable variation; it does not make any scale track reliably.

---

#Ask about these

Before stepping on the scale, or before worrying about a number:

1. Is this number measured, or calculated? Apart from weight, essentially all of them are calculated.
2. I differ from yesterday — is that me, or my hydration? What the scale directly measures is impedance, and body water is derived from it; when fluid status changes, impedance and the numbers derived from it may change too.
3. Which BMI classification am I using? Your scale may use WHO's general classification (25/30), while Hong Kong's report uses the Asian classification (23/25).
4. I want to know about visceral fat or central obesity — have I measured my waist? The official thresholds are women ≥ 80 cm and men ≥ 90 cm, with a tape measure.
5. I want to know about osteoporosis — the scale cannot answer. What it reports is an estimate of bone mineral weight; osteoporosis is diagnosed on bone density.

If you have an implanted electronic medical device such as a pacemaker or respirator, the Consumer Council states you should not use a smart scale to measure body composition.

---

#Frequently asked questions

• How does a smart scale "know" my body fat percentage? It does not measure fat directly. The Consumer Council explains that most smart scales use bioelectrical impedance analysis, using resistance data and preset equations to measure or calculate body fat percentage, water content, muscle mass, bone mass, basal metabolic rate and so on; different smart scales may use different equations, so the figures may differ.
• Two scales of different makes disagree — which is right? This article makes no judgement. The Consumer Council notes that equations differ between scales, so the figures may differ. A study comparing 15 bioelectrical impedance devices (14 consumer-grade, 1 research-grade) with a laboratory four-compartment model likewise found constant error against the reference method ranging from −3.5 to +11.7 percentage points.
• So is a scale useless? The same study found precision error on repeat measurement of 0.0% to 0.49%, but the authors note that 5 foot-to-foot devices had a least significant change of only 0% to 0.11% — "suspiciously high" reliability that may reflect an algorithm suppressing differences on repeat measurement within a short period. For tracking change over 12 to 16 weeks, constant error was −0.4% to 1.3% and Lin's CCC 0.37 to 0.78. The authors conclude that, although performance varied widely, select models (particularly octapolar and select foot-to-foot devices) may hold potential utility for tracking body composition over time.
• Why does my body fat percentage change after drinking water? Bioelectrical impedance measures the impedance of current passing through the body, and water conducts differently from fat. The manufacturer's own page states that body water is "constantly changing" and is affected by alcohol, flu and menstruation among other things. This article cites no figure for "how many percentage points a glass of water changes".
• What is "metabolic age"? On the manufacturer's page opened for this article, metabolic age is the result of comparing your basal metabolic rate with the average of your age group — a comparison, not a measurement. This article has not reviewed every brand's definition, and therefore makes no statement about whether a corresponding reference measurement method exists.
• Can a scale's "bone mass" tell me whether I have osteoporosis? No. The manufacturer defines it as the estimated weight of bone mineral in the body, produced by a statistical calculation based on studies showing a close correlation between bone weight and lean mass (the manufacturer does not publish the equation, nor state that lean mass is the sole input). Osteoporosis is diagnosed on bone mineral density, a different measurement. A comparative study also found that existing bioelectrical impedance equations show substantial bias in estimating bone mineral content (mean difference up to 0.46 kg), and that DXA remains the gold standard.
• Does Hong Kong have an official standard for body fat percentage? This article checked all 81 pages of the Report of Population Health Survey 2020-22 (Part II) (2023): the terms for body fat, fat percentage, muscle mass, bone density, basal metabolic rate, metabolic age and bioelectrical impedance appear zero times. The indicators the report uses are BMI (Asian classification 23.0/25.0), waist circumference (women ≥ 80 cm, men ≥ 90 cm, International Diabetes Federation Asian classification) and waist-to-hip ratio (women ≥ 0.85, men ≥ 0.90, World Health Organization). This article makes a statement about that one report only, and has not checked other official documents.
• Who should not use a smart scale? The Consumer Council states that people with implanted electronic medical devices such as a pacemaker or respirator should not use a smart scale to measure body composition, so that its low-frequency current does not affect the device's operation.

Also in this series:

• Which check-ups have official intervals: “How often should you have a body check? Which items have an official interval, and why packages do not 2026”
• How a "biological age" test is calculated: “Is a biological age test worth doing? A full breakdown of retests differing by up to 9 years, the prices available in Hong Kong, and the disclaimers the sellers write themselves 2026”
• What the evidence says on weight loss methods: “Does intermittent fasting work for weight loss? A full breakdown of how the control group changes the answer, the blank after 24 weeks, and where three sources disagree 2026”

---

#What this article does not claim

• It does not claim that all smart scales measure only weight and impedance. It refers to the body composition figures derived from impedance; individual models have other sensing functions (the manufacturer's page opened for this article lists resting heart rate).
• It derives no individual error range from the constant error span across the 15 devices. Constant error is a model's mean bias; interpretation at the individual level requires that model's standard error of the estimate, proportional bias and limits of agreement.
• It does not claim that a change in hydration moves every output in step. None of the sources opened makes that statement, and BMI is not computed through impedance.
• It makes no statement about whether a corresponding reference measurement method exists for metabolic age. It cites one manufacturer's definition only.
• It rates, compares and recommends no brand or model of smart scale. The manufacturer named appears only as the source of its own definitions, not as a recommendation.
• It opened only one manufacturer's explanatory page. All "what the manufacturer says" quotations represent that manufacturer's definitions of its own readings, not an industry standard; other manufacturers' definitions may differ.
• It cites no numeric range for any "visceral fat level". The page opened describes what visceral fat is but prints no numeric levels.
• It cites no figure for "how much a glass of water changes the reading". No such trial was opened; on hydration affecting readings, the article cites only the manufacturer's own statement that body water is constantly changing, with no number attached.
• It does not apply the standing research-grade analyser's 3% correction factor to household scales. That study compared a research/clinical-grade instrument with DXA, and its scope is stated in the text.
• It does not claim that no Hong Kong official document has ever set these thresholds. It checked one document — the Report of Population Health Survey 2020-22 (Part II) — and makes a statement about that report only.
• It cites no "healthy range" for body fat percentage. The Population Health Survey report prints no body fat threshold; the body water range mentioned on the manufacturer's page is marked by the manufacturer itself as a guideline only, and is not cited here as a threshold.
• It does not deal with recommendations on osteoporosis screening. That a scale cannot answer on bone density is explained in the text; official screening arrangements are covered in this site's article on check-up intervals.
• It offers no medical, nutritional or weight-loss advice. Every figure is quoted, with its source, study population and scope; what your reading means and whether to act on it is for a doctor to decide.
• The three studies cited and the manufacturer's page are all English-language sources. This English version quotes them from their English originals.

---

Provenance: compiled from a manufacturer's own definitions, a statutory body's consumer information, an official population health survey and peer-reviewed literature; every figure is marked with its source, study population and the date the document itself prints.

Quotations from Chinese-language official documents and press materials appear in our own English translation; the original wording governs.

#資料來源 (Sources)

• TANITA Europe, "Understanding your measurements" (page carries no date of its own; English only): https://tanita.eu/understanding-your-measurements (retrieved 2026-08-30)
• Consumer Council, "Smart scales: more than just weighing" (page prints 2017.06): https://www.consumer.org.hk/ws_chi/news/specials/2017/Keep-fit-pound.html (retrieved 2026-08-30)
• Department of Health, Centre for Health Protection, Report of Population Health Survey 2020-22 (Part II) (title page prints "Department of Health, Centre for Health Protection, Non-Communicable Disease Branch, 2023"): https://www.chp.gov.hk/files/pdf/dh_phs_2020-22_part_2_report_chi_rectified.pdf; parent page: https://www.chp.gov.hk/tc/features/37474.html (retrieved 2026-08-30)
• Assessing the reliability and cross-sectional and longitudinal validity of fifteen bioelectrical impedance analysis devices. Br J Nutr (document prints issue date 2023 Sep 14; English only): https://pmc.ncbi.nlm.nih.gov/articles/PMC10404482/ (retrieved 2026-08-30)
• High precision but systematic offset in a standing bioelectrical impedance analysis (BIA) compared with dual-energy X-ray absorptiometry (DXA) (document prints collection date 2022 Dec; English only): https://pmc.ncbi.nlm.nih.gov/articles/PMC9813632/ (retrieved 2026-08-30)
• Bioelectrical impedance analysis of bone mineral content based on dual-energy X-ray absorptiometry: evaluation of age-stratified optimized models. Scientific Reports (English only): https://www.nature.com/articles/s41598-025-08304-8 (retrieved 2026-08-30)

Compiled by the editorial team from a manufacturer's own statements, a statutory body's consumer information, an official survey report and academic literature; every figure is attributed to its source. This is health information, not medical advice, and not a product recommendation.