Part 3 of 3 in Paceline's Wearable Accuracy Series. Part 1: Heart rate and HRV accuracy · Part 2: Sleep tracking and readiness scores

Steps are usually in the right ballpark. VO2max estimates can be off by several points for any one person. Calorie burn is the least accurate thing your wearable reports, with no brand consistently getting it right. Blood oxygen and skin temperature are useful for spotting changes over time but aren't medical measurements. This final part of the series goes through the evidence and then pulls the whole series together: why readings from different devices can't be compared directly, and what to do about it now.

How accurate are VO2max estimates on Apple Watch and Garmin?

VO2max (the most oxygen your body can use during all-out effort) is one of the strongest markers of long-term health. Measuring it properly means a maximal treadmill or bike test with a mask. Watches estimate it from heart rate, pace and personal details.

• Meta-analysis (14 studies): devices that estimate VO2max from resting data "significantly overestimated" it. Devices using exercise data had near-zero average bias (−0.09 mL/kg/min) but individual limits of agreement from −9.92 to +9.74 mL/kg/min. The authors called the individual-level error "large" (Molina-Garcia et al., 2022).
• Umbrella review: wearables overestimated VO2max by about 15.24% in resting tests and 9.83% in exercise tests (Doherty et al., 2024).
• Apple Watch (Series 5 to 9 era, 30 adults): underestimated VO2max by 6.07 mL/kg/min, MAPE 13.31%, limits of agreement −6.11 to +18.26 (Lambe et al., 2025; device series per Lambe et al., 2026).
• Apple Watch Series 7 (19 adults): MAPE 15.79%, with poor reliability (ICC 0.47). The watch tended to overestimate VO2max in less-fit people and underestimate it in very fit people (Caserman et al., 2024).
• Garmin fēnix 6 (19 adults): MAPE 7.05%, concordance 0.73, meeting the authors' validity criteria (Carrier et al., 2025).
• Older trackers (2019 study, five devices): MAPE above 10% for everything except the Garmin Forerunner 920XT's VO2max estimate (Passler et al., 2019).

These are small studies, and the devices in them are one to several generations old. Still, the pattern holds: the number on your watch is a reasonable trend line, but two watches can put the same person several points apart.

Why are wearable calorie counts so inaccurate?

Energy expenditure isn't measured directly. It's modeled from movement, heart rate and your profile, and the models differ between brands and activities.

• Meta-analysis (60 studies, 104 effect sizes): accuracy varied by activity, with very high heterogeneity (I² above 75%) for many devices. Adding heart rate to accelerometry helped (O'Driscoll et al., 2020, Br J Sports Med).
• Stanford, seven devices, 60 people: "No device achieved an error in EE below 20 percent." Apple Watch had the lowest error and Samsung Gear S2 the highest (Shcherbina et al., 2017).
• Systematic review, 65 articles: calorie MAPE was above 30% for all brands (Germini et al., 2022). Another review of 158 publications concluded "no brand was accurate" for energy expenditure (Fuller et al., 2020).
• Apple Watch: every subgroup in one meta-analysis exceeded the 10% validity threshold for calories (Choe & Kang, 2025). In another, MAPE ranged from 9.71% (running) to 151.66% (walking), and five of the eight calorie studies tested a Series 2 or older (Lambe et al., 2026).

Researchers in the INTERLIVE network say validation methods for calorie estimates "lack rigour", which makes device-to-device comparison hard (Argent et al., 2022). In plain terms: don't eat back your watch's calories, and never compare your burn with a friend's on a different brand.

How accurate are wearable step counts?

Steps are the most mature metric, but not perfect:

• Wearables mostly undercount, with MAPE ranging from about −9% to 12% (Doherty et al., 2024).
• In lab settings, Fitbit, Apple Watch and Samsung "appeared to measure steps accurately" (Fuller et al., 2020). Fitbit Charge models stayed under 25% MAPE across 20 studies (Germini et al., 2022).
• Independent Apple Watch step validation has mostly covered the first-generation and Series 1 models (Lambe et al., 2026).

Where you wear the device, your walking speed and daily tasks like pushing a stroller all shift counts, which is why INTERLIVE published a separate step-count validation protocol (Johnston et al., 2021).

Is smartwatch blood oxygen (SpO2) accurate?

At normal oxygen levels, wearables are close on average but vary for individuals. When oxygen drops, which is when it matters, they struggle:

• Umbrella review: mean absolute difference up to 2.0% (Doherty et al., 2024).
• Apple Watch meta-analysis: bias near zero (−0.04%) but limits of agreement from −4.00 to +3.94% (Lambe et al., 2026).
• Controlled low-oxygen testing: Apple Watch Series 7 readings fell within 2% of arterial blood values 32.14% of the time, versus 49.03% for a Masimo fingertip oximeter. Both overestimated oxygen, and both exceeded FDA error thresholds below 88% (Jiang et al., 2026).
• Garmin fēnix 6: SpO2 concordance was just 0.10 (MAPE 4.29%) (Carrier et al., 2025).

There's also an equity issue. Even medical pulse oximeters missed dangerously low oxygen (arterial saturation below 88% despite a reading of 92–96%) in 11.7% of measurements in Black patients versus 3.6% in White patients in one hospital cohort (Sjoding et al., 2020, NEJM). Wearable SpO2 hasn't been tested across diverse skin tones nearly enough.

What about skin temperature?

The evidence here is thin. Rings and watches report temperature as a deviation from your baseline, not an absolute body temperature, and they measure different spots (finger versus wrist).

• In 22 women, Oura's nocturnal finger temperature rose 0.30 °C in the luteal phase of the menstrual cycle and correlated moderately with oral temperature (r = 0.563) (Maijala et al., 2019). Two authors were Oura employees.
• The TemPredict study showed illness-related temperature rises were detectable in 50 people with COVID-19, which the authors described as a feasibility result (Smarr et al., 2020).
• Apple Watch wrist temperature is listed among metrics "yet to be validated" (Lambe et al., 2026).

VO2max, calorie, step and SpO2 accuracy: what the studies found

Why wearable data needs normalization: the series in one picture

Across all three parts, the same three problems keep coming up.

1. Sensors and models differ in accuracy. HRV error ranged from about 6% to 16% between brands (Part 1). Sleep-stage agreement ranged from 50% to 65% (Part 2). VO2max and calorie estimates vary by several points and tens of percent.

2. Brands define the "same" metric differently. Garmin defines resting HR as the lowest 30-minute average in 24 hours. Others use a sleep average. HRV is RMSSD on some platforms and SDNN on another, measured over the whole night, the first 4 hours, or weighted to late deep sleep. Readiness scores combine these inputs with undisclosed weights.

3. The evidence trails the hardware. Only about 11% of released wearables have been validated for even one metric, and existing studies are 3.5% of what's needed (Doherty et al., 2024). Many headline studies test models that are several years old. Independent data on Coros and Ultrahuman is close to nonexistent.

So a reading from one brand and a reading from another aren't directly comparable, even when both devices are "accurate." Any cross-device score, leaderboard or benchmark needs to normalize readings, adjusting for the device, the metric's definition and the hardware generation, before it puts people side by side. The authors of the umbrella review call for "standardised validation protocols and collaborative industry partnerships" for the same reason.

That's the direction Paceline is building in: fair, consistent tracking for people who use whatever device they already own, not only people who buy one particular brand.

What you should do right now

1. Pick one device and stick with it for anything you want to trend.
2. Compare yourself to yourself. Weekly averages on the same device beat single readings and cross-brand comparisons.
3. Treat calories as a rough guide. Don't eat back workout calories number for number.
4. Use VO2max as a trend. Watch direction over months. Get a lab test if you need a precise number.
5. Keep your profile accurate. Age, height, weight and max HR all feed the estimates.
6. Record outdoor runs with GPS and a chest strap when you want a better VO2max estimate. Exercise-based estimates beat resting ones.
7. Don't use smartwatch SpO2 for medical decisions. If you feel short of breath or unwell, seek care.
8. Read temperature as a change from your own baseline, never as a thermometer reading.
9. When you switch brands, start a new baseline. Your old numbers and new numbers aren't on the same scale.

Where gear can help

For better exercise-based estimates and steadier training data, these third-party partner products are in the Paceline Marketplace:

• Garmin Forerunner 570 or Garmin Venu 4: GPS watches that estimate VO2max from outdoor workouts.
• Garmin HRM 600: a chest strap for more reliable heart rate during the workouts that feed those estimates.
• Withings ScanWatch 2: a hybrid watch with SpO2 and temperature tracking. Read those as trends, per the evidence above.
• For long or hot sessions: CURE Electrolyte Drink Mix, Watermelon or Thorne Daily Electrolytes.

None of these makes readings comparable across brands. Better inputs make your own trend more trustworthy.

Keep your trend consistent, whatever you wear

Your wearable is most valuable as a consistent record of you. The Paceline app connects the device you already own and rewards steady week-to-week habits, so you don't have to switch brands to take part.

Download the Paceline app

Read the full series: Part 1: Heart rate and HRV · Part 2: Sleep and readiness scores

FAQ

Is Apple Watch or Garmin VO2max more accurate?

In separate small studies, Garmin fēnix 6 had a MAPE of 7.05% and Apple Watch 13.31–15.79%. The studies used different protocols and older models, so there's no definitive head-to-head answer.

Why does my watch say I burned 600 calories and my friend's says 400?

Different brands model calories differently, and studies find errors of 20–30% or more across devices. Neither number is precise.

Can I trust my smartwatch's blood oxygen reading?

Use it as a wellness trend only. It's least accurate at low oxygen levels, which is when accuracy matters most.

Sources

1. Molina-Garcia P, et al. Validity of estimating the maximal oxygen consumption by consumer wearables: a systematic review with meta-analysis and expert statement of the INTERLIVE Network. Sports Med. 2022;52(7):1577-1597. PubMed 35072942
2. Lambe R, O'Grady B, Baldwin M, Doherty C. Investigating the accuracy of Apple Watch VO2 max measurements: a validation study. PLoS One. 2025;20(5):e0323741. PubMed 40373042
3. Caserman P, et al. Assessing the accuracy of smartwatch-based estimation of maximum oxygen uptake using the Apple Watch Series 7. JMIR Biomed Eng. 2024;9:e59459. PubMed 39083800
4. Carrier B, Marten Chaves S, Navalta JW. Validation of aerobic capacity (VO2max) and pulse oximetry in wearable technology. Sensors. 2025;25(1):275. PubMed 39797066
5. Passler S, et al. Validity of wrist-worn activity trackers for estimating VO2max and energy expenditure. Int J Environ Res Public Health. 2019;16(17):3037. PubMed 31443347
6. O'Driscoll R, et al. How well do activity monitors estimate energy expenditure? A systematic review and meta-analysis. Br J Sports Med. 2020;54(6):332-340. PubMed 30194221
7. Shcherbina A, et al. Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. J Pers Med. 2017;7(2):3. PubMed 28538708
8. Germini F, et al. Accuracy and acceptability of wrist-wearable activity-tracking devices: systematic review of the literature. J Med Internet Res. 2022;24(1):e30791. PubMed 35060915
9. Fuller D, et al. Reliability and validity of commercially available wearable devices for measuring steps, energy expenditure, and heart rate: systematic review. JMIR Mhealth Uhealth. 2020;8(9):e18694. PubMed 32897239
10. Choe JP, Kang M. Apple Watch accuracy in monitoring health metrics: a systematic review and meta-analysis. Physiol Meas. 2025;46(4). PubMed 40199339
11. Lambe R, et al. The accuracy of Apple Watch measurements: a living systematic review and meta-analysis. npj Digit Med. 2026;9:63. PubMed 41513748
12. Doherty C, et al. Keeping pace with wearables: a living umbrella review. Sports Med. 2024;54(11):2907-2926. PubMed 39080098
13. Argent R, et al. Recommendations for determining the validity of consumer wearables and smartphones for the estimation of energy expenditure (INTERLIVE). Sports Med. 2022;52(8):1817-1832. PubMed 35260991
14. Johnston W, et al. Recommendations for determining the validity of consumer wearable and smartphone step count (INTERLIVE). Br J Sports Med. 2021;55(14):780-793. PubMed 33361276
15. Jiang Y, et al. Performance of wearable pulse oximetry during controlled hypoxia induction. JMIR Form Res. 2026;10:e85253. PubMed 41894520
16. Sjoding MW, et al. Racial bias in pulse oximetry measurement. N Engl J Med. 2020;383(25):2477-2478. PubMed 33326721
17. Maijala A, et al. Nocturnal finger skin temperature in menstrual cycle tracking: ambulatory pilot study using a wearable Oura ring. BMC Womens Health. 2019;19:150. PubMed 31783840
18. Smarr BL, et al. Feasibility of continuous fever monitoring using wearable devices. Sci Rep. 2020;10:21640. PubMed 33318528

This article is for educational purposes only and isn't medical advice. Consumer wearables aren't medical devices. Don't use SpO2, temperature or heart readings to diagnose or manage a medical condition. Talk to a qualified clinician.