OptimalBod
The body, measured.
A sensor-fusion engine that separates real fat loss from the water, glycogen, and day-to-day noise that fools every scale — and learns what your body actually does with the calories you eat.
A sensor-fusion engine that separates real fat loss from the water, glycogen, and day-to-day noise that fools every scale — and learns what your body actually does with the calories you eat.
OptimalBod pairs your daily nutrition with four independent measurements of your body — scale weight, bioimpedance body-fat, silhouette-photo geometry, and the energy you burn — anchored to periodic DEXA scans.
From these it produces a single, continuously-updated estimate of your fat mass and lean mass over time, each with a genuine confidence band. And because it pairs what you eat with the physics of energy balance, it does something no ordinary tracker can: it backs out your true daily energy expenditure from the gap between what you logged and what your body actually did.
The headline is your filtered body-fat with its margin — never a bare number. Trends draws the credible band tightening at each DEXA scan; Energy backs out your true TDEE; Capture keeps your photos on-device; Data shows the app's own leave-one-out accuracy.
Your morning weight swings a kilo or two from water and glycogen alone — a scale literally cannot tell fat loss from a dry Tuesday. Every other body-composition app has only observations, so the best it can do is average them and hope.
OptimalBod is built the other way around. It starts from a physical model — energy in minus energy out predicts tomorrow's fat mass — and uses each day's measurements to correct that prediction rather than trust it blindly. That one inversion is what makes adaptive TDEE, lean-mass alerts, and plateau diagnosis possible. It's the difference between a bathroom scale and the filter that keeps a spacecraft on course.
Here's the quiet truth about body-composition tracking: Apple Health already has almost everything OptimalBod does. The same Withings scale, the same bioimpedance, the same active energy and sleep from your Watch. Dozens of apps read those exact numbers.
What none of them have is what you ate.
That single omission is why they can only ever smooth and display. Energy balance — calories in minus calories out — is the engine of body change, and calories in come from one place: your nutrition. Feed that into the model and the whole system changes character. It stops merely recording your weight and starts predicting it — then correcting that prediction against the scale, the bioimpedance, the photos.
Nutrition is what turns OptimalBod from a mirror into a model. It's the term that makes true TDEE observable, that lets the filter tell a real deficit from a logging slip, that powers the required-intake solver. Without it, you have a very handsome chart of the past. With it, you have an instrument that tells you what to do next.
| Calorie apps | Smart scales | DEXA | OptimalBod | |
|---|---|---|---|---|
| Daily tracking | ✓ | ✓ | quarterly | ✓ |
| Separates fat from water & glycogen | ✕ | ✕ | one snapshot | ✓ |
| Feeds nutrition into the model | logs only | ✕ | ✕ | ✓ |
| Learns your real TDEE | preset formula | ✕ | ✕ | ✓ |
| Adapts as your body changes | ✕ | ✕ | ✕ | ✓ |
| Warns on muscle loss | ✕ | ✕ | if you compare | ✓ |
| Shows its uncertainty | ✕ | ✕ | ±1–2% | ✓ |
| Runs fully on-device & private | cloud | cloud | clinic | ✓ |
A calorie app logs input and guesses output from a static equation. A smart scale gives you a hydration-noisy number with no memory. A DEXA is accurate but arrives once a quarter with no trend in between. OptimalBod fuses all of them — and gets sharper every time a scan lands.
At the core is a six-state Extended Kalman Filter stepping once per day, tracking not just what you weigh but the hidden quantities underneath it:
Each day's energy surplus or deficit is split between fat and lean by the Forbes partition, using the real energy densities of tissue (ρF = 9,417 and ρL = 1,816 kcal/kg):
The G state is modeled as an Ornstein–Uhlenbeck process — mean-reverting water and glycogen — which is exactly what lets a daily weigh-in become usable signal instead of being misread as fat. The headline term, ε, absorbs the steady gap between logged and actual energy balance; invert it and you have your true TDEE.
The photo channel is deterministic geometry, not a vision AI — circumferences from Ramanujan's ellipse approximation feeding the U.S. Navy equation, computed entirely on-device, so the same photo always returns the same number:
Each new observation nudges the estimate; a Rauch–Tung–Striebel smoother runs backward whenever a DEXA scan arrives, so a single scan today retroactively sharpens months of past history:
And sleep isn't ignored — it enters as covariates that shift the fat/lean partition and tell the filter to distrust the morning scale after a bad night, when hydration is most volatile.
Fat mass, lean mass, and body-fat % over time — always with a credible band, never a bare number pretending to certainty.
Actual daily expenditure backed out from your own data and reported honestly — wide at first, tightening as evidence accumulates.
A date range to reach your target body-fat %, drawn from the filter's uncertainty — not a false-precision single day.
The daily calories to land on target by a chosen date, recomputed each week as your BMR falls with your fat mass.
Catches muscle loss the moment it outpaces the model — the failure a scale hides completely, and the single most useful warning the app makes.
Tells you when the scale stalls but fat is still falling, and names water retention when the data points there.
A real n-of-1 experiment correlating your rolling sleep debt with your fat-loss rate and partition — correlation, honestly bounded.
Safe-deficit clamps, a BMR intake floor, and essential-fat warnings — enforced in the model, not as dismissible copy.