Wearable makers face growing regulatory scrutiny of their data practices — the FTC's 2023-2024 enforcement actions over sensitive health and location data included settlements with health-app and data-broker firms — but the accuracy of the training numbers athletes rely on remains essentially self-policed, validated by each manufacturer's own studies. A 2020 Stanford study of seven wrist-worn devices found energy-expenditure errors of 27 to 93 percent, while heart-rate accuracy was considerably better. For training-minded readers, the practical news is that regulation is tightening privacy faster than measurement.
What has regulators' attention?
Data flows, not data quality. The FTC's Health Breach Notification Rule actions and its settlements over sensitive-data sharing established that health-adjacent app data gets legal protection — the 2023 GoodRx order and 2024's location-data cases set the pattern, per the agency's own releases. The EU's GDPR already treated fitness data as sensitive. Nothing equivalent forces a sports watch's calorie count to be accurate within any bound — no NIST-style standard exists for consumer wearables, a gap academic reviewers in sports science have noted repeatedly.
How inaccurate are the numbers, really?
It depends which number. The Stanford group's 2020 paper in Nature's partner journal Medicine & Science in Sports & Exercise measured 60 volunteers across seven devices: heart rate erred by about 5 percent, while energy expenditure erred far more — the 27 to 93 percent range, with no device reliably close. The detail other coverage skipped: error grew with darker skin tones for optical heart rate in some earlier studies, a documented equity gap in the technology that Stanford's group and others have flagged. Sleep staging and VO2 estimates carry similar company-by-company uncertainty; validation studies, where they exist, are usually the maker's own.
What does it change for training?
A hierarchy of trust. Heart rate and pace — directly measured signals — are sound enough for training decisions. Derived estimates — calories, sleep stages, stress scores, recovery readiness — are model outputs wearing the confidence of measurements; treat trends within one device as useful and cross-device comparisons as rough. A cyclist making weight decisions from calorie math, or a runner interpreting a readiness score as gospel, is leaning on the weakest numbers on the wrist.
What the record establishes: privacy regulation is catching wearables while accuracy remains vendor-validated, and the largest documented errors sit in the most-consumed metric. What remains unknown is whether any regulator takes up measurement claims — no rulemaking on the table promises it.
