Why fitness trackers overestimate calories burned

Here's what the research actually found, why the math breaks down, and what to do about it.

What the research actually found

In 2017, a team at Stanford put seven popular wrist-worn devices — Apple Watch, Fitbit Surge, Microsoft Band, Mio Alpha 2, PulseOn, Samsung Gear S2, and Basis Peak — through sitting, walking, running, and cycling while measuring real energy expenditure with indirect calorimetry (the lab-grade method of measuring the oxygen you actually consume). Every device was tested against the same ground truth.

The heart rate sensors did fine — most devices came within 5% of the real heart rate during cycling. The calorie estimates did not. Not one device got within 20% of the real number. The best performer, the Apple Watch, still ran about 27% off. The worst, PulseOn, was off by 93% — about as useful as a coin flip (Shcherbina et al., Journal of Personalized Medicine, 2017).

A more recent study backs this up with newer hardware. Researchers at Florida International University put four current watches — Apple Watch Series 8, Fitbit Sense 2, Samsung Galaxy Watch 5, and Garmin Forerunner 955 — on 58 adults riding a stationary bike through alternating moderate and vigorous intervals, checked against a COSMED K5 metabolic analyzer (the clinical gold standard). Typical errors landed in the 15–25% range, with some watches running well past that. Apple was again the most accurate; Garmin and Samsung overestimated the most (FIU, published in PLOS ONE).

Nine years and several hardware generations apart, two independent labs landed on the same conclusion: the number on your wrist is a guess, and it skews high.

Why the math breaks down

A tracker can't measure calories directly — nobody's smartwatch has a tiny respirometer. What it actually has is a heart rate sensor and an accelerometer, and it runs those two signals through a formula to estimate energy expenditure. That formula is built from population averages: age, sex, height, weight, and a model of how heart rate typically relates to oxygen consumption during exercise.

The heart rate part is genuinely solid — that's just pulse detection, and the hardware is good at it. The conversion from "pulse went up" to "you burned X calories" is where it falls apart, because that relationship isn't fixed. It depends on your actual fitness level, your muscle mass, how efficiently your body moves, and plenty else the formula never sees. Two people with the identical heart rate during the same workout can burn meaningfully different amounts of energy, and the watch has no way to tell them apart.

Why it gets worse with harder workouts and higher body fat

Both studies found the error wasn't random — it had a direction. The Stanford study saw the highest relative error during low-intensity activity like sitting, where the absolute numbers are small but the percentage error balloons. The FIU study found something more specific and more useful: error grew as the wearer's body-fat percentage increased, because the algorithms assume a body composition that doesn't match everyone wearing the device.

Resistance training and interval-style workouts are a particular trouble spot — heart rate lags behind effort during short, intense bursts, and the device ends up averaging over noise it wasn't designed to handle. If your workout doesn't look like a steady jog, treat the number with extra suspicion.

What this means for your calorie math

The practical risk isn't that the number is wrong — it's what people do with it. If your tracker says you burned 500 calories and you eat back all 500, and the real number was closer to 300, you've just erased 40% of whatever deficit you were trying to create. Do that most days and a stalled scale stops being a mystery.

Use the number as a trend, not a budget line: today's workout vs. last week's, roughly harder or easier, not an exact figure to bank calories against. If you do want to eat back exercise calories, eating back half of what the tracker reports is a reasonable hedge against the kind of overestimation these studies found — still a guess, just a less generous one.

How Bocco treats this number

I don't try to out-guess your metabolism either — nobody's cracked that problem, trackers included. What I did build Bocco to do is keep the number you can verify — the food you ate — accurate and correctable, instead of asking you to trust a black box on both ends. If a meal estimate looks off, you say so in one sentence and the math redoes itself. That's a much smaller, much more honest problem than estimating energy expenditure from a wrist sensor.