The higher the body fat percentage, the greater the margin of error? An inconvenient truth revealed by smartwatch research

The higher the body fat percentage, the greater the margin of error? An inconvenient truth revealed by smartwatch research

The "500kcal" displayed on your wrist is an estimate, not a measurement

The moment you finish exercising, your smartwatch displays "500kcal burned." The sense of accomplishment from seeing your efforts quantified might lead some to think, "Now I can have dessert." However, that 500kcal is not a value measured directly within your body. It is an estimate calculated by the manufacturer's proprietary algorithm, using factors like heart rate, arm movement, age, gender, height, and weight.

This distinction is not just a matter of semantics. If you view the display merely as a guideline for exercise, some margin of error is acceptable. However, if you use it as a "budget" to determine calorie intake during weight loss, that error could directly translate into discrepancies in meal portions. New research suggests that these discrepancies not only vary by device but could also widen based on body fat percentage.


Comparing four popular models with "respiratory gas"

The research team at Florida International University tested four devices: Apple Watch Series 8, Fitbit Sense 2, Samsung Galaxy Watch 5, and Garmin Forerunner 955. The subjects were 58 Hispanic adults aged 18 to 50, with an average age of 23. Participants pedaled a recumbent bike, alternating between moderate and high intensity every two minutes for a total of 10 minutes. Before and after this, there were five-minute rest and recovery periods, during which the watches' exercise modes were active for a total of 20 minutes.

The benchmark was the COSMED K5 metabolic measurement device, which analyzes oxygen and carbon dioxide in breath to determine energy expenditure. Unlike indirect information from the wrist, this method evaluates metabolism through respiratory gas and is used as a comparison standard in laboratory settings.

The research team examined the "bias," which is the difference between the watch's value and the benchmark, the "absolute error," which measures the magnitude of the difference regardless of sign, and the "absolute percent error," which is the error rate relative to the benchmark. Without distinguishing these, one might overlook the phenomenon where "the average seems correct, but individual results vary widely."


Garmin and Samsung overestimate, Apple shows relatively small discrepancies

When comparing measurements that met data quality standards, the average bias was +21.6kcal for Apple, +68.6kcal for Garmin, and +56.8kcal for Samsung. This means that under these conditions, all three devices generally displayed higher values than actual, with Garmin and Samsung showing particularly large overestimations.

Fitbit's average bias was +3.1kcal, which might seem the most accurate at first glance. However, this figure alone does not justify concluding that "Fitbit is accurate." Fitbit recorded seven instances of unnatural values exceeding 450% of the benchmark, accounting for about 13% of the trials for this device. The average of +3.1kcal was calculated after excluding these outliers, and before exclusion, the average bias expanded to +128.6kcal. There were also measurements where no value was recorded.

Additionally, the absolute error of model estimates at the participants' average body fat percentage was approximately 25.2kcal for Apple, 18.8kcal for Fitbit, 47.9kcal for Garmin, and 41.0kcal for Samsung. The smaller value for Fitbit should be noted as it results from an analysis excluding significant outliers. Across the study, the median typical absolute percent error was reported to be roughly 15-25% depending on the model.

However, this is not a ranking of "Apple is always accurate" or "Garmin is always about 70kcal too high." These results were obtained under limited conditions: the specific models, short-duration cycling exercise, and the attributes of the participants. Results could differ for outdoor running, strength training, daily life, or devices from different generations.


Higher body fat percentage led to increased errors across all models

The most noteworthy finding of this study is that as body fat percentage increases, the error in calorie estimation also increases for all four brands. The rate of error increase varied by brand, with percent error expanding more rapidly for Fitbit and Garmin compared to Apple, and absolute error increases being more pronounced for Garmin and Samsung.

Why body fat percentage affects the results cannot be definitively determined from this experiment alone. Many smartwatches combine optical heart rate sensors, which read blood flow changes by shining light on the skin, with accelerometers. The thickness of subcutaneous tissue, blood flow, movement of soft tissue during exercise, and the snugness of the wristband could influence the signal. Moreover, if the data used to train and adjust the algorithm lacks sufficient diversity in body types, larger errors could occur for specific body shapes.

Meanwhile, within the Fitzpatrick skin type classification range of III to V examined in this study, no clear relationship between skin type and error was confirmed. However, only four participants had the darkest type V skin, and types I, II, and VI were not included. Thus, it cannot be generalized that "skin color does not affect accuracy."


The headline "Harmful to Dieting" is not directly proven by the study

There is an important point to note. This study measured the accuracy of estimated energy expenditure during exercise, but it did not track whether people using smartwatches actually gained weight or failed to lose weight. Therefore, it is a step too far to assert that "smartwatches harm dieting."

Nonetheless, the potential risk can be understood. For example, if the watch overestimates calorie expenditure and the user compensates by eating the same amount, the expected calorie deficit could be smaller. If errors consistently accumulate in the same direction, it could lead to a situation where "the calculations suggest weight loss, but the scale does not move." The study's lead researcher also warns that treating the display as an exact figure could result in discrepancies of several hundred kcal per week, turning an expected deficit into a surplus.

Conversely, fearing errors and drastically reducing food intake is also dangerous. Energy is necessary for recovery and sustained exercise, and significant intake restrictions are not justified simply because the display is inaccurate. The issue is not the use of the smartwatch itself, but treating the estimates as precise accounting figures.


On social media, "untrustworthy" and "suits me" coexist

This study was just announced at the end of July 2026, and as far as can be confirmed through public searches, reactions to the study itself on social media are still limited. However, when looking at related user posts, the discussion broadly divides into three categories.

 

The first category is the pragmatic view of "using the absolute value as a guideline." In a Reddit post questioning the calorie display of the Apple Watch, there were voices advocating ignoring the watch's numbers and focusing on personal perception, or using a chest strap in conjunction while viewing the watch as a baseline. This pragmatic approach aligns with the researchers' suggestions.

The second category is confusion over the large differences when switching brands. In a thread where a user reported a 300-400kcal difference in daily display after switching from Apple Watch to Garmin, some said "Garmin is closer to my actual measurements," while others shared opposite experiences, saying "it's about 500kcal lower than my total expenditure." The significant individual differences make it clear that a single example from social media cannot determine superiority.

The third category points out that "the definitions of the numbers being compared may differ." In a post reporting a 778kcal difference between Apple Watch and Fitbit, responses suggested checking whether active calories and total calories burned, including basal metabolism, were being confused. Even the same term "calories burned" can cover different scopes depending on the app or device. A screenshot comparison between brands does not necessarily equate to a precision comparison.

Additionally, some users report that "when cross-referencing meal records and weight changes, the watch's estimates were quite accurate for me." This is entirely possible, but individual experiences are not controlled evaluations. While social media can provide insights into the diversity of errors and user practices, it is neither a representative survey of the population nor a basis for product rankings. The existence of anecdotes that seem to contradict the experiment results underscores the importance of focusing on long-term trends within individuals rather than absolute values.


Five rules for leveraging smartwatches in dieting

First, do not directly equate displayed calories with "extra allowance to eat." Consuming 100% of the calories displayed from exercise each time is disadvantageous for those who consistently experience overestimation. However, there is not enough evidence to suggest a uniform percentage to deduct, so extreme corrections should be avoided.

Second, keep your profile updated with current weight, height, age, etc., and select a workout mode that matches your exercise. Ensure the band is snug enough not to slip on your wrist, and keep the sensors clean. Manufacturers also explain that fit, skin blood flow, cold, tattoos, and exercises involving continuous wrist bending can affect heart rate measurements. While updates and calibrations can improve some aspects, they do not eliminate the inherent limitations of the algorithm.

Third, do not simply compare numbers between different devices. Continue using the same device, in the same workout mode, under similar conditions, and observe relative changes like "was my activity level higher than last week?" Research papers also state that even if absolute accuracy is low, tracking changes within an individual is not invalidated.

Fourth, make weight loss decisions based on multiple factors such as weight trends over several weeks, meal records, hunger levels, and exercise performance. Daily weight can fluctuate due to water, so it's better to look at moving averages or weekly averages rather than daily changes. The watch's numbers should be just one piece of the decision-making puzzle.

Fifth, individuals with chronic conditions like diabetes, those managing energy intake for medical reasons, or those with a history of eating disorders should not adjust their diet based solely on wearable displays and should consult with a doctor or registered dietitian. Calorie expenditure displays are not diagnostic values from medical devices.


Remember that this is a small-scale, single-condition study

The strength of the study lies in equipping the same participants with all four devices and directly comparing them with respiratory gas analysis. However, there are many limitations. The study was small-scale with 58 participants, limited to Hispanic adults with an average age of 23. The exercise was a single session on a recumbent bike, and it did not examine daily life or long-term use. The simultaneous wearing of multiple devices could have affected band position and snugness.

Furthermore, even if devices have mechanisms to learn and readjust to users over time, this cannot be evaluated in a single experiment. While outlier exclusion was planned in advance, specific numerical criteria were confirmed during data review, and the impact of exclusion was particularly significant for Fitbit. The devices tested were from around the 2022 generation, and the results cannot be directly applied to the latest models or current software.

Nevertheless, past studies have repeatedly reported that "while heart rate can be relatively measured, energy expenditure is difficult." The current results are meaningful in showing that not only has the issue not been resolved, but also that errors may be biased by body composition.


Conclusion—Change the role of numbers, don't discard them

The value of smartwatches is not solely in the absolute calorie numbers. They are useful tools for tracking exercise time, checking heart rate zones, becoming aware of excessive sitting, and providing motivation for continuity. Therefore, the solution is not to stop wearing the watch but to change the role of the numbers.

Consider calorie expenditure as an "indicator of the general direction of activity" rather than an "amount to be used for settling accounts." Track long-term trends with the same device and verify reality with actual measurements of diet and weight. The most practical way to wisely use smartwatches is not to follow the numbers displayed on your wrist, but to use them as one piece of information.


Source URL

Related Articles