In this guide
How Bioelectrical Impedance Analysis Estimates Body Fat
Quick answer: Bioelectrical impedance analysis (BIA) sends a tiny alternating current through the body and measures how much the body resists it. Muscle holds water and conducts well; fat holds almost no water and resists. From that resistance, plus your height, weight and age, an equation estimates total body water, then fat-free mass, then fat mass. The number is an estimate built on assumptions about hydration — never a direct measurement of fat, and never as precise as a DEXA scan.
Why this matters now
On 12 November 2026, Withings Health Solutions announced Body Pro 3, a cellular-connected scale for weight-management programmes that adds a retractable handle so it can measure body composition segment by segment — each arm, each leg and the torso — instead of one whole-body number. Withings reports that in a 2025 validation study against DEXA, its readings correlated 0.99 for fat mass and 0.98 for lean and muscle mass. The announcement matters for students because it shows where the technology is heading: the interesting engineering is no longer the weighing, it is the impedance measurement and what you honestly do with it.
This guide explains the physics behind that measurement, why the numbers move from morning to evening, and how a student build translates the headline device into a safe lab-scale version. The companion student project is the Withings Body Pro 3-Inspired Smart Scale Project with Body Composition Analysis and Segmental Fat Tracking, which builds the scale with an AD5933 impedance-analyser chip, an HX711 load-cell ADC and an ESP32 dashboard.
The core idea: water conducts, fat resists
Body tissue is not one material. Muscle is roughly three-quarters water with dissolved electrolytes, so it conducts electricity well. Fat tissue holds very little water, so it resists current. Bone and skin add their own resistance. If you know how much current you pushed in and how much voltage came back, you know the impedance — and impedance is mostly a story about how much water the current met on its way through.
That is the whole trick. BIA never sees fat directly. It measures an electrical property, converts it into an estimate of total body water, and then works backwards through a chain of assumptions: total body water implies fat-free mass (because fat-free tissue has a fairly stable water fraction), and fat mass is simply body weight minus fat-free mass. Every step after the impedance reading is arithmetic on assumptions, which is why the result is always an estimate.
How the measurement actually runs
A BIA scale works in four stages:
- Weigh first. Four load cells under the platform measure body weight precisely. Weight is the one number in the whole chain that is measured directly, not estimated.
- Apply a small alternating current. The scale sends a low-current alternating signal through the electrodes — the standard single frequency in consumer scales is 50 kHz. Alternating current is used because direct current would polarise the electrodes and irritate the skin; at 50 kHz the current passes through tissue without any sensation.
- Measure impedance. The analyser measures the voltage that develops across the body for that known current, giving impedance: a resistance part (mostly water and electrolytes) and a reactance part (mostly cell membranes acting as tiny capacitors).
- Run the equations. The device combines impedance with the person's height, weight, age and sex in a published bioelectrical impedance equation set, and outputs fat mass, muscle mass and a visceral-fat index. Height matters because the current travels a longer path in a taller person; age and sex matter because the equations were fitted to population data.
Frequency choice is not arbitrary. At low frequencies the current flows mainly around cells, through the extracellular fluid. At higher frequencies it crosses cell membranes and samples intracellular water too. A single 50 kHz reading sits in the middle and is the standard compromise for consumer devices; multi-frequency devices sweep several frequencies to separate the two water compartments, at the cost of a more complex analyser.
Segmental measurement: why a handle changes the answer
A basic scale has four electrodes under the feet and measures one path: foot to foot, which mostly samples the legs. The reading for the arms and torso is then inferred, not measured. Body Pro 3's retractable handle adds four hand electrodes, so the device can probe several paths — hand to foot, hand to hand, foot to foot — and estimate each segment separately. That is the segmental reading: fat and muscle for the left arm, right arm, torso, left leg and right leg, instead of one whole-body number.
The student build copies exactly this idea with a fold-out handle carrying four stainless-steel electrodes, plus four on the platform. The honest limit stays the same: each segment value comes from the same impedance paths and carries the same uncertainty as the whole-body figure. A segmental display looks more precise than a single number, but it is the same measurement cut into five slices.
Why the same person gets different numbers
Anyone who has owned a body-composition scale has watched the fat percentage move by a point or two between morning and evening. That movement is real measurement behaviour, not a fault in the scale:
- Hydration dominates. Drink a litre of water and the body conducts better; the scale reads lower impedance and reports less fat. Dehydration after exercise does the opposite. Total body water is the quantity actually being sensed, so anything that changes it changes the answer.
- Food and exercise shift water between compartments. A meal, a workout, or a hot shower redistributes water and changes skin temperature, all of which move the impedance reading.
- Skin and electrode contact matter. Dry hands grip the handle electrodes well; wet or sweaty palms lower the contact resistance and shift the reading. Socks add an insulating layer the sweep cannot see through, which is why the reading procedure insists on bare feet.
- Time of day is a variable, not a detail. The honest way to use a BIA scale is to compare readings taken at the same time each morning, under the same conditions, and watch the trend over weeks — not to chase a single reading.
This is also why a BIA scale should never be presented as a medical device. It estimates, it does not diagnose, and it cannot measure blood pressure or detect hypertension. The reported Body Pro 3 hypertension feature analyses weigh-in data patterns, does not measure blood pressure, is pending regulatory clearance and is not available — even the headline device makes no blood-pressure claim.
BIA vs DEXA: what the 0.99 correlation actually means
DEXA (dual-energy X-ray absorptiometry) is the clinical reference for body composition. It passes two low-dose X-ray beams through the body and measures how much each tissue type absorbs, which separates bone, lean tissue and fat far more directly than impedance ever can. A validation study that reports a 0.99 correlation between a BIA device and DEXA is reporting agreement across a study population, not a guarantee for any one person on any one morning.
Correlation is also not the same as agreement on absolute values. Two methods can track each other closely across a group while still differing by a few percentage points for an individual. When Withings reports 0.99 for fat mass and 0.98 for lean and muscle mass from its 2025 validation study, that is a strong population-level result for their device and their equations — and it says nothing about what a student-built scale with a different analyser, different electrodes and different equations will achieve. The student build states this plainly: its figures are estimates, never DEXA-level measurements.
The student build: how the headline device becomes a lab project
The companion project translates Body Pro 3 into hardware a student can actually build and defend in a review:
- Weighing: four 50 kg half-bridge load cells under a tempered-glass platform, read by an HX711 24-bit ADC at a gain of 128 and 10 samples per second.
- Impedance: an AD5933 impedance-analyser chip (12-bit, 1 MSPS ADC, frequency generator to 100 kHz, I2C interface) runs the sweep at 50 kHz through 8 stainless-steel electrodes.
- Segmental paths: four foot electrodes on the platform plus four hand electrodes on a fold-out handle, mirroring the reported retractable handle.
- Profiles and trends: an ESP32 keeps up to 6 user profiles, serves a local web dashboard over Wi-Fi, and draws a trend projection fitted by simple linear regression to each profile's own weigh-in history — labelled a projection, never a clinical prediction.
- Calibration you can repeat: weight is calibrated with a known standard weight; the impedance path is calibrated with a precision 1 kΩ resistor measured through the AD5933 before any body measurement.
Two builder details decide whether the readings are usable at all. First, the HX711 is read at 10 samples per second while the sweep runs, not 80: the faster setting widens the ADC filter and lets more supply noise into the weight reading while the AD5933 is exciting the electrodes. Second, the handle electrodes must be gripped with dry hands, and the platform needs bare feet — wet palms and socks both change the contact resistance the sweep sees. Both details are in the project report as builder notes, because they are the difference between a repeatable reading and noise.
What to check before trusting any body-composition number
- Is weight measured or estimated? Weight should come from load cells. If a device cannot show you a stable, repeatable weight, its composition figures are built on sand.
- Is the figure labelled an estimate? Any honest BIA product or project labels fat and muscle figures as estimates. A page that presents them as measurements is overstating the method.
- Was it calibrated? A known weight for the scale and a known resistor for the impedance path are the minimum. Without both, there is no chain from the reading back to a reference.
- Are the reading conditions fixed? Same time of day, same hydration state, bare feet, dry hands. A trend across weeks under fixed conditions is informative; a single reading is not.
- Does it claim blood pressure or diagnosis? A scale that measures weight and impedance cannot measure blood pressure. Any such claim is a reason to stop trusting the page.
Frequently asked questions
How accurate is bioelectrical impedance analysis?
Weight is precise because load cells measure it directly. Fat mass, muscle mass and visceral-fat figures are estimates computed from impedance, and they move with hydration, skin contact and time of day. Compared against DEXA on a population, a well-validated commercial device can correlate closely — Withings reports 0.99 for fat mass in its 2025 validation study — but that is their device and their equations, not a property of the method in general, and not a promise for any single reading.
Why does my body-fat percentage change during the day?
Because the method senses water, and your water distribution changes. Drinking, eating, exercise, sweating and showering all move water between compartments and change skin temperature, which shifts impedance. Compare readings taken at the same time each morning rather than chasing individual numbers.
Do I need to hold a handle for a body-composition reading?
Only if you want segmental values. Foot-only electrodes measure mainly the legs and infer the rest. Hand electrodes add arm and torso paths, so the device can estimate each segment separately. The handle must be gripped with dry hands for the full sweep, or the contact resistance changes the reading.
Can a smart scale measure blood pressure?
No. A scale measures weight and impedance; nothing in that hardware senses blood pressure. Withings' reported Hypertension Notification Feature analyses weigh-in data patterns, does not measure blood pressure, is pending regulatory clearance and is not available. Any scale that claims to measure blood pressure from a weigh-in is misrepresenting the method.
What frequency does BIA use, and why?
The standard single frequency in consumer scales is 50 kHz. At that frequency the alternating current passes through tissue without sensation, and it samples a useful mix of extracellular and intracellular water. Lower frequencies stay mostly outside cells; higher frequencies cross cell membranes. Multi-frequency devices sweep several frequencies to separate the compartments, but a single 50 kHz reading is the standard student-build choice.
Is a BIA scale a medical device?
No — not the student build, and not a consumer smart scale. It is an estimation tool for tracking trends, not a diagnostic instrument. It must never be used for diagnosis, treatment or clinical decisions, and its figures must always be labelled as estimates.
Limitations
- BIA estimates body composition; it never measures fat directly, and it never reaches DEXA-level accuracy for an individual.
- Readings depend on hydration, food intake, exercise, skin contact and time of day — fixed reading conditions are part of the method, not optional.
- Segmental values carry the same uncertainty as the whole-body estimate; cutting one measurement into five segments does not make it five times more precise.
- The AD5933 measures 100 Ω to 1 kΩ only with additional circuitry, so body paths below 1 kΩ are outside the student build's range.
- Nothing in this guide or the companion project measures blood pressure, diagnoses any condition, or replaces clinical assessment.
- The Withings figures quoted here (DEXA correlations, GLP-1 context, availability) are reported figures from the 12 November 2026 announcement, not measurements of the student build.
Need help with the measurement side of your project?
If your project needs weighing, impedance or sensor measurements and you are unsure which chip or method fits, message us on WhatsApp with the sensor details, the accuracy you need, and your deadline. We will reply with a quotation stating the scope, price, and delivery timeline for the build.
Related reading: Withings Body Pro 3-Inspired Smart Scale Project · How EchoMotion Tracks Your Whole Body with Ultrasound from a Smartwatch · How Smart Home Hubs Recognise Faces and Give Every Person Their Own Screen · ESP32 Projects: How the ESP32 Works and 7 Project Ideas · more in the IoT & Embedded branch hub.