One budget fitness band overstated calories by up to 57%. Another was off by 13%

The calorie number on a fitness band is the figure most people act on and the one that is hardest to verify. A study published this month in PLOS One put four inexpensive, widely sold wrist bands against laboratory indirect calorimetry and found errors ranging from tolerable to very large [s1].

The test

Twenty Chinese women who exercised three times a week or less each wore two of the four bands simultaneously, randomly assigned to each wrist, while cycling on an ergometer at four fixed workloads: 30 W, 40 W, 50 W and 60 W [s1]. Energy expenditure was measured at the same time by indirect calorimetry using a CORTEX METAMAX 3B, which the authors treat as the gold standard [s1].

The four devices were the HONOR Band 7, HUAWEI Band 8, XIAOMI Smart Band 8 and KEEP Smart Band B4 Lite [s1].

Fixed-workload cycling is an unusually clean test. Unlike walking or running, where stride and terrain vary, ergometer cycling at a set wattage produces a known mechanical demand, so any disagreement between the band and the calorimeter is device error rather than variation in what the person did.

The results

Across all four workloads, the XIAOMI Smart Band 8 and the KEEP Smart Band B4 Lite significantly overestimated energy expenditure relative to the calorimeter (all p < 0.001) [s1]. The HONOR Band 7 and HUAWEI Band 8 did not [s1].

The magnitude, expressed as mean absolute percentage error across the four loads: 49.5% to 57.4% for the KEEP band, 30.5% to 41.0% for the XIAOMI band, 15.0% to 23.0% for the HONOR band, and 12.5% to 18.6% for the HUAWEI band [s1]. A lower value means the estimate sits closer to the measured value [s1].

The spread between the best and worst device is roughly fourfold. A band reporting 400 kilocalories with a 55% error is reporting something a long way from what was actually expended.

Bland-Altman analysis showed a consistent pattern across all four devices: positive bias increased as energy expenditure rose, most clearly in the XIAOMI and KEEP bands [s1]. In other words, the harder the effort, the more the devices overshot. That direction matters, because it means the error is largest during exactly the sessions people are most likely to count.

What the study does not establish

The sample is 20 untrained women, and the authors name the small, homogeneous sample and occasional missing data across load levels as limitations [s1]. Nothing here transfers automatically to men, to trained athletes, to other activities, or to newer firmware on the same hardware — device algorithms are updated frequently and validation studies necessarily test a snapshot.

Nor does the study test step counts or heart rate, which are measured more directly and generally validate better than energy expenditure. Calorie burn is a derived estimate: the device infers it from motion and heart rate through a proprietary model, which is why it is the metric most prone to divergence between brands.

The authors' own framing is a caution about a specific use — they suggest care in using low-price smartwatches for energy balance management in untrained female populations [s1]. This article does not offer advice on weight management or exercise; the finding is about what the number on the screen represents.

Where this sits in the wider literature

A review published earlier in the month in npj Cardiovascular Health surveys consumer-grade wearables for clinicians and researchers who are not device specialists, describing the common sensors, the health parameters they can measure, and the state of research into the validity and accuracy of those measurements [s2]. Its framing is that these devices offer a genuine opportunity for understanding public health trends, developing risk stratification tools and monitoring interventions [s2] — while treating accuracy as an open question that varies by parameter and by device rather than a solved one.

The PLOS One study is one data point inside that open question, and it lands on the parameter with the weakest track record.

What to watch

Whether manufacturers publish validation data for energy expenditure at all. Step count and heart rate accuracy are commonly reported; the derived calorie figure, which is the number most consumers use, generally is not — and this study suggests the gap between devices on that metric is wide enough to matter.

Sources

• Validity of four low-cost smartwatches in estimating energy expenditure during cycling in Chinese untrained women — PLOS One, 2025-09-23
• A guide to consumer-grade wearables in cardiovascular clinical care and population health for non-experts — npj Cardiovascular Health, 2025-09-02