Smartwatches overestimate calories burned during exercise

Smartwatches can provide useful information about physical activity, but their estimates of calories burned during exercise should be interpreted with caution. Researchers at Florida International University (FIU) found that several popular fitness watches consistently overestimated energy expenditure compared with clinical laboratory measurements. Accuracy also appeared to vary according to users’ body composition.

The study, published in PLOS One, compared four widely used smartwatches with laboratory equipment for measuring energy expenditure. According to the researchers, inaccuracies could be particularly relevant for people who use calorie estimates to manage their weight.

Four popular watches tested

The researchers recruited 58 adults who completed a controlled workout on a recumbent exercise bike. The session alternated between moderate and vigorous intensity. Participants wore an Apple Watch Series 8, Fitbit Sense 2, Samsung Galaxy Watch 5 or Garmin Forerunner 955. Their actual energy expenditure was simultaneously measured using a COSMED K5 metabolic analyser, equipment used in clinical laboratory settings.

Typical errors were approximately 15 to 25 percent, although average deviations were considerably higher for some of the devices. Of the watches included in the comparison, the Apple Watch produced the most accurate estimates, while the Garmin and Samsung devices showed the largest overestimations.

The researchers were unable to draw firm conclusions about the Fitbit Sense 2. During testing, the device occasionally generated implausibly low estimates, including a reading of only one calorie for an entire workout. In other cases, no result was produced. Much of the Fitbit data was therefore excluded from the analysis.

Body fat associated with larger errors

The study also identified an association between body fat percentage and measurement accuracy. The higher a participant’s body fat percentage, the greater the error in estimated calorie expenditure tended to be. Lead author Jason Kostrna, associate professor of kinesiology and exercise science at FIU, said the researchers do not yet know exactly why body composition affects the estimates. One possible explanation is the algorithms manufacturers use to translate measurements from wearable sensors into estimates of energy expenditure.

The datasets used by manufacturers to develop these algorithms are generally not known to the researchers. More diverse development and validation data could potentially help improve the accuracy of future devices.

The finding could have practical implications for people using smartwatches as part of weight-management programmes. According to Kostrna, accumulating relatively small errors across multiple workouts could result in a difference of hundreds of calories per week. Someone relying on these estimates could therefore believe they are maintaining a calorie deficit when their actual energy balance is different.

Useful, but not a clinical measurement

The findings do not mean smartwatches have no value for monitoring physical activity. The researchers recommend viewing calorie estimates as approximate indicators rather than precise measurements of energy expenditure. Users could, for example, interpret estimates conservatively rather than automatically compensating for all reported calories through food intake.

At the same time, the researchers caution against responding to potential overestimation by excessively reducing calorie intake. Adequate nutrition remains necessary for recovery and sustained physical activity. Other wearable measurements may provide more actionable information. Heart rate and time spent within different heart-rate zones, for example, can help users monitor exercise intensity and changes in endurance over time.

Kostrna and colleagues are now extending their research to strength training. Short, intense periods of exertion may present different challenges for algorithms than cycling and other continuous forms of exercise. The next studies should provide more insight into how reliably popular wearables estimate energy expenditure across different types of physical activity.