Smartwatches Overestimate Calories, Worse at Higher Body Fat

IT TOOK ten minutes of pedalling, a rubber mask, and four watches strapped to one wrist after another to expose what a decade of marketing had not. The gap between what a smartwatch claims you burned and what your body actually burned traces mainly to how each brand’s proprietary formula converts heart rate and motion into calories, tuned on leaner test populations, and has surprisingly little to do with the skin tone sensors are usually blamed for. A new trial out of Florida International University (FIU) put four of the most popular fitness trackers on the same wrists as a clinical grade metabolic analyzer and let the machines disagree. The disagreement was not small, and it was not evenly spread.

Wearables like these already sit on countless wrists, feeding decisions about how much to eat after a run or whether a diet is working. The same energy expenditure blind spot has turned up in other independent lab comparisons of popular trackers, even on devices that measured heart rate itself quite well. When the number on the screen overstates the real cost by dozens of kilocalories, the decision built on top of it is off too.

Under the strap, a smartwatch never actually measures a calorie. It watches a beam of light bounce off blood vessels near the skin, called photoplethysmography, and pairs the resulting heart rate signal with an accelerometer and a demographic based formula to guess at effort. That formula is proprietary, built somewhere you cannot see, and calibrated in ways companies do not have to disclose. When the wrist it is guessing about does not resemble the wrists in that calibration set, the guess drifts.

Researchers already suspected two obvious culprits for a bad guess: how dark a person’s skin is, since melanin can absorb some of the light the sensor relies on, and how much body fat sits between the sensor and the vessel it is trying to read. The Florida team set out to test both at once rather than assume which one mattered more.

Four Watches, Four Different Numbers

58 Hispanic adults with Fitzpatrick skin types III to V rode a recumbent bike through alternating stretches of moderate and vigorous effort while wearing an Apple Watch Series 8, a Fitbit Sense 2, a Samsung Galaxy Watch 5 and a Garmin Forerunner 955, all four at once. A COSMED K5 unit, the kind of equipment used in exercise physiology labs rather than living rooms, measured their real oxygen use breath by breath, the criterion every watch was judged against. Every watch overestimated calories burned on average, but not by equal amounts: Garmin’s bias worked out to roughly 69 kcal too many per session, Samsung’s to roughly 57, and Apple’s to a comparatively modest 22.

Jason Kostrna, the study’s lead author and an associate professor of kinesiology and exercise science at FIU, puts it thus: “You can’t treat the calories burned number it gives you as an accurate number, because it’s not.” The stakes are not abstract: “You could be hundreds of calories off each week and easily end up in a calorie surplus when you think you’re in a calorie deficit.”

The Body Fat Blind Spot

The pattern sharpened once the team looked at body fat percentage rather than the device alone. As body fat rose, every watch’s error grew with it, though not at the same rate: Garmin’s and Fitbit’s errors climbed fastest, while Apple’s stayed comparatively flat, an interaction the statistics called significant. One obvious suspect, ironically, turned out not to matter here: Fitzpatrick skin type, the classification the study was partly designed to test, showed no measurable effect on accuracy at all, a null result the press release covering this paper never mentions. Fitbit’s own numbers hide a complication. Averaged the ordinary way, its bias looked the smallest and least alarming of the four, but that average excluded seven sessions, about 13% of its trials, where the watch produced an implausible number, and putting those back in pushes Fitbit’s true mean overestimate to roughly 129 kcal, the worst of any device tested.

None of this makes a smartwatch useless for tracking effort. It means the calorie total on the screen behaves more like a ceiling than a receipt, especially for the people using it to manage their weight who also carry more body fat, the group the study found least well served.

The trial has limits of its own. It tested one exercise modality, a recumbent bike, and one population, Hispanic adults with medium to deep skin tones, so whether the same gaps show up on a run, or in a lighter skinned rider, is still an open question.

Kostrna does not have a tidy explanation for why body fat trips up the algorithms, but he has a guess: the formulas were likely built and tested on leaner volunteers, and a heavier body simply falls outside the range they were calibrated for. “When the companies are developing those algorithms, we don’t know where the test data is coming from,” he said. “That’s why I think it’s beneficial for the companies to see this type of data, so they can refine their products over time.” He is already running the follow up, testing how the same four watches handle strength training instead of cycling.

Kostrna’s advice for now sits closer to a coping strategy than a fix: watch the trend line, not any single number, and let how you feel guide the day more than the total on your wrist. “Track your own trends over time,” he said. “If it becomes too much, don’t fixate on it.”

FAQ