Consumer sleep trackers detect sleep reliably but struggle to recognize wake: in a 2025 validation of six popular wrist devices against polysomnography, every device detected more than 90% of sleep epochs, yet correctly identified only 29.39% to 52.15% of wake epochs (Sleep Advances, Schyvens et al. 2025). Epoch-by-epoch agreement with the lab standard ranged from a Cohen’s kappa of 0.21 (Garmin Vivosmart 4) to 0.53 (Apple Watch Series 8), the band researchers describe as fair to moderate. The best-performing devices reach 75.0% to 76.3% epoch agreement across four sleep stages, but in patients with sleep complaints agreement can fall to a kappa of 0.07. Meanwhile adoption keeps climbing: 48% of US adults have now used a sleep tracking device, up from 35% in 2023 (American Academy of Sleep Medicine, 2025). We aggregated data from the Journal of Clinical Sleep Medicine, Sleep Advances, Sleep Medicine, the Journal of Sleep Research, the American Academy of Sleep Medicine, the US Food and Drug Administration, Apple’s clinical validation paper and the other primary sources listed in the methodology. For the broader context on how much Americans actually sleep, see our sleep statistics.
TL;DR
- Six wrist trackers each detected more than 90% of sleep epochs, but wake specificity was 29.39% to 52.15% (Sleep Advances, 2025).
- Epoch agreement kappa ranged from 0.21 (Garmin Vivosmart 4) to 0.53 (Apple Watch Series 8) (Sleep Advances, 2025).
- Oura Ring Gen3 showed 91.7% to 91.8% sleep-wake accuracy across 421,045 epochs in 96 adults (Sleep Medicine, 2024).
- Four-stage agreement: Oura 76.3%, Apple Watch 75.0%, Fitbit 70.9% in an Oura-funded study (Sensors, 2024).
- Apple Watch Series 8 underestimated deep sleep by 43 minutes and overestimated light sleep by 45 minutes (Sensors, 2024).
- A meta-analysis of 24 studies and 798 participants found wrist devices differed significantly from PSG on total sleep time, sleep efficiency, latency and wake (Journal of Clinical Sleep Medicine, 2025).
- In patients with sleep complaints, agreement kappa fell to 0.12 (Withings) and 0.07 (Fitbit) (Journal of Sleep Research, 2025).
- Device and self-reported sleep agreed 82.4% of the time in good sleepers but only 39.4% in insomnia patients (Sleep Medicine, 2026).
- Samsung’s Sleep Apnea Feature received an FDA De Novo grant on February 6, 2024 (FDA, DEN230041).
- Apple’s sleep apnea notifications showed 66.3% sensitivity and 98.5% specificity in 1,448 participants (Apple validation study, 2024).
- 48% of US adults have used a sleep tracking device, up from 35% in 2023 (AASM survey, 2025).
- 55% of adults change their behavior based on sleep tracker data (AASM survey, 2025).
1. Sleep vs Wake: Excellent at Sleep, Weak at Wake
The headline accuracy figure most brands quote, sleep detection, is the easy part. Most of a night is sleep, so a device that leans toward scoring ‘asleep’ will look accurate overall. All six devices in a 2025 validation detected more than 90% of sleep epochs, but specificity, the share of true wake epochs they caught, ranged from just 29.39% to 52.15% (Sleep Advances, Schyvens et al. 2025). In practice that means a tracker can miss half or more of the time a user lies awake, which is exactly the time people with poor sleep care about.
The pattern is not new. A 2021 study in the journal Sleep tested seven consumer devices in 34 healthy young adults over three nights and found epoch sensitivity of at least 0.93 for every device but specificity of only 0.18 to 0.54 (Sleep, Chinoy et al. 2021). Ring-based devices have narrowed the gap: the Oura Ring Gen3 reached 73.0% to 74.6% specificity in a 96-participant study, still well below its 94.4% to 94.5% sensitivity (Sleep Medicine, Svensson et al. 2024).
| Metric | Value | Source |
|---|---|---|
| Sleep epoch sensitivity, six wrist devices | More than 90% for all devices | Sleep Advances, Schyvens et al. 2025 |
| Wake specificity, six wrist devices | 29.39% (Garmin Vivosmart 4) to 52.15% (Apple Watch Series 8) | Sleep Advances, Schyvens et al. 2025 |
| Epoch agreement kappa, six wrist devices | 0.21 (Garmin) to 0.53 (Apple); Fitbit Sense 0.42, Fitbit Charge 5 0.41, Whoop 4.0 0.37, Withings Scanwatch 0.22 | Sleep Advances, Schyvens et al. 2025 |
| Total sleep time overestimation | Fitbit Sense 6.31 min to Withings Scanwatch 39.87 min (Whoop 4.0 24.46, Apple 19.60, Garmin 38.44) | Sleep Advances, Schyvens et al. 2025 |
| Oura Ring Gen3 sleep-wake | Sensitivity 94.4% to 94.5%, specificity 73.0% to 74.6%, accuracy 91.7% to 91.8% | Sleep Medicine, Svensson et al. 2024 |
| Oura Ring Gen3 agreement | PABAK 0.83 to 0.84; reliability 94.8% | Sleep Medicine, Svensson et al. 2024 |
| Two-stage epoch agreement (sleep vs wake) | Oura 92.0% (kappa 0.60), Fitbit 91% (kappa 0.52), Apple Watch 93% (kappa 0.60) | Sensors, Robbins et al. 2024 |
| Seven consumer devices, sleep-wake | Sensitivity at least 0.93; specificity 0.18 to 0.54 | Sleep, Chinoy et al. 2021 |
Context note: the six-device study enrolled 62 adults (52 men, 10 women; mean age 46.0), and not every participant wore every device (Apple Watch n = 20, Garmin n = 25). The 2021 Sleep study is most recent available data for its seven-device comparison.
2. Sleep Stage Agreement: The Mid-70s Ceiling
Splitting sleep into light, deep and REM is where consumer trackers earn or lose credibility, and even the best results top out in the mid-70s. In a 35-participant inpatient study at Brigham and Women’s Hospital, four-stage epoch agreement with PSG was 76.3% for the Oura Ring Gen3 (kappa 0.65), 75.0% for the Apple Watch Series 8 (kappa 0.60) and 70.9% for the Fitbit Sense 2 (kappa 0.55) (Sensors, Robbins et al. 2024). That study was funded by Oura Ring Inc., a conflict readers should weigh.
The minute-level biases are large enough to change how a night looks. The Apple Watch underestimated deep sleep by 43 minutes and overestimated light sleep by 45 minutes in the Brigham study, and a 2026 home-PSG study of healthy adults aged 18 to 45 found the Apple Watch Series 7 underestimated light sleep by 73.8 minutes and overestimated REM by 30.3 minutes, while the Fitbit Charge 5 overestimated light sleep by 57.7 minutes (Sleep Medicine, Alhejaili et al. 2026). The authors judged agreement ‘generally weak with wide’ limits of agreement and suited to longitudinal self-monitoring, not clinical assessment.
| Metric | Value | Source |
|---|---|---|
| Four-stage epoch agreement | Oura Gen3 76.3% (kappa 0.65), Apple Watch S8 75.0% (kappa 0.60), Fitbit Sense 2 70.9% (kappa 0.55) | Sensors, Robbins et al. 2024 |
| Stage sensitivity range, three devices | 50% to 86% | Sensors, Robbins et al. 2024 |
| Deep sleep sensitivity | Oura 79.5%, Fitbit 61.7%, Apple 50.5% | Sensors, Robbins et al. 2024 |
| Apple Watch Series 8 stage bias | Deep -43 min, light +45 min, wake -7 min | Sensors, Robbins et al. 2024 |
| Fitbit Sense 2 stage bias | Light +18 min, deep -15 min | Sensors, Robbins et al. 2024 |
| Oura Ring Gen3 staging accuracy | 75.5% (light) to 90.6% (REM); REM underestimated by 4.1 to 5.6 min | Sleep Medicine, Svensson et al. 2024 |
| Apple Watch Series 7 bias (home PSG) | Light -73.8 min, REM +30.3 min | Sleep Medicine, Alhejaili et al. 2026 |
| Fitbit Charge 5 and Polar Vantage M2 bias | Fitbit light +57.7 min; Polar REM +32.3 min | Sleep Medicine, Alhejaili et al. 2026 |
Context note: the 2026 Alhejaili study allocated each participant to one device (Fitbit n = 20, Apple Watch n = 19, Polar n = 15), so per-device samples are small. Stage results are not comparable across studies that use different device generations and firmware.
3. What the Meta-Analyses Say
Single validations vary with firmware, population and lab protocol, so the pooled evidence matters more. A Journal of Clinical Sleep Medicine meta-analysis of 24 studies and 798 participants found wrist-worn trackers differed significantly from PSG on total sleep time (mean difference -16.854 minutes), sleep efficiency, sleep latency and wake after sleep onset (Journal of Clinical Sleep Medicine, 2025). Its verdict: popular, but ‘not as reliable as polysomnography’ for key parameters.
A 2026 Behavioral Sleep Medicine review of 16 studies (10 pooled) reached the same direction of error: wearables overestimated total sleep time and sleep efficiency and underestimated wake after sleep onset, with the Oura Ring Gen3 closest to PSG on some metrics (Behavioral Sleep Medicine, Agostinho et al. 2026). The most favorable pooled result is ring-specific: an OTO Open meta-analysis of 6 studies and 388 participants found no statistically significant difference between the Oura Ring and medical-grade studies on seven parameters (OTO Open, 2025).
| Metric | Value | Source |
|---|---|---|
| Wrist devices meta-analysis scope | 24 studies, 798 participants | Journal of Clinical Sleep Medicine, 2025 |
| Total sleep time vs PSG | Mean difference -16.854 min (95% CI -26.332 to -7.375) | Journal of Clinical Sleep Medicine, 2025 |
| Sleep efficiency vs PSG | Mean difference -4.691 (95% CI -7.079 to -2.302) | Journal of Clinical Sleep Medicine, 2025 |
| Sleep latency and WASO vs PSG | Latency 2.574 (95% CI 0.606 to 4.542); WASO 13.255 (95% CI 4.522 to 21.988) | Journal of Clinical Sleep Medicine, 2025 |
| Healthy-adult wearable review | 16 studies included, 10 in meta-analysis | Behavioral Sleep Medicine, 2026 |
| Oura Ring vs medical-grade studies | 6 studies, 388 participants (mean age 64.7) | OTO Open, 2025 |
| Oura pooled differences | Total sleep time -2.97 min (95% CI -10.27 to 4.33); deep sleep 1.39 min; REM -3.89 min | OTO Open, 2025 |
Context note: the JCSM meta-analysis found no significant WASO difference for the Fitbit subgroup, and no significant latency difference for non-Fitbit devices. Pooled analyses mix device generations released years apart, so they describe the category rather than any model on sale today. Heart-rate sensing, which feeds these staging algorithms, is covered in our fitness tracker heart rate statistics.
4. Clinical Populations: Where Accuracy Breaks Down
Most validations recruit healthy young adults, the group least likely to need a sleep tracker’s help. In 65 patients referred to a sleep clinic, overall epoch agreement kappa was 0.12 for a Withings device and 0.07 for a Fitbit, and the Fitbit’s kappa was 0.00 in insomnia and -0.15 in hypersomnia (Journal of Sleep Research, 2025). A kappa near zero means agreement no better than chance. The authors concluded commercial trackers ‘should not replace actigraphy and/or PSG’ in this group.
Perception gaps widen too. A 2026 Sleep Medicine systematic review of 5 studies and 2,006 participants found device and self-reported sleep quality agreed 82.4% of the time in good sleepers but only 39.4% in insomnia patients, with wearables explaining just 2.5% to 16.2% of the variance in subjective sleep (Sleep Medicine, 2026). In a 108-patient hospital study, turning frequency and sleep posture significantly drove a Huawei Band 9’s total sleep time error (Nature and Science of Sleep, 2025).
| Metric | Value | Source |
|---|---|---|
| Epoch agreement kappa, sleep clinic patients | Withings 0.12; Fitbit 0.07 | Journal of Sleep Research, 2025 |
| Fitbit kappa by indication | Insomnia 0.00; hypersomnia -0.15 | Journal of Sleep Research, 2025 |
| Total sleep time overestimation, patients (median) | Withings 28 min; Fitbit 10 min | Journal of Sleep Research, 2025 |
| Device vs self-reported concordance | 82.4% good sleepers vs 39.4% insomnia patients | Sleep Medicine, 2026 |
| Variance in subjective sleep explained by wearables | 2.5% to 16.2% (5 studies, 2,006 participants) | Sleep Medicine, 2026 |
| Fitbit Charge 4 in Huntington’s disease | Stage time ICC 0.79 to 0.96; wake sensitivity 68.6% | Journal of Clinical Sleep Medicine, 2024 |
| Total sleep time ICC vs PSG, young adults | Fitbit Versa 2 0.86, Withings Mat 0.85, Fitbit Inspire HR 0.84, Oura Gen2 0.77, Garmin Vivosmart 4 0.38 | Sensors, Kainec et al. 2024 |
Context note: the Huntington’s disease study had only 10 participants, and the Kainec study tested 53 adults aged 18 to 30 without epoch-level data, so its ICCs reflect nightly totals only (Sensors, Kainec et al. 2024).
5. Sleep Apnea Features and Regulation
The regulatory line is clear: tracking sleep is a lifestyle feature, assessing disease risk needs FDA authorization. The FDA granted Samsung’s over-the-counter Sleep Apnea Feature a De Novo classification on February 6, 2024, creating a new device type to ‘assess risk of sleep apnea’ (FDA De Novo DEN230041). The American Academy of Sleep Medicine’s position remains that consumer sleep technology ‘must be FDA cleared and rigorously tested against current gold standards’ if it is intended to diagnose or treat (AASM Position Statement, JCSM 2018).
Apple’s notification feature shows the tradeoff of a screening tool tuned to avoid false alarms. In its clinical validation of 1,448 participants, sensitivity was 66.3% and specificity 98.5%, but sensitivity was 43.4% in moderate cases versus 89.1% in severe ones (Apple, Sleep Apnea Notifications on Apple Watch, September 2024). Apple estimates sleep apnea affects more than 1 billion people worldwide, mostly undiagnosed (Apple Newsroom, September 2024); our obstructive sleep apnea prevalence statistics cover that burden in detail.
| Metric | Value | Source |
|---|---|---|
| Samsung Sleep Apnea Feature | FDA De Novo granted February 6, 2024 (DEN230041) | US FDA De Novo database |
| Apple validation study | 1,499 enrolled, 1,448 completed | Apple validation paper, September 2024 |
| Apple sensitivity and specificity | 66.3% (95% CI 62.2% to 70.3%) and 98.5% (95% CI 98.0% to 99.0%) | Apple validation paper, September 2024 |
| Apple sensitivity by severity | Moderate 43.4%; severe 89.1% | Apple validation paper, September 2024 |
| Apple algorithm development data | 3,936 nights from 2,160 participants (design), 7,220 nights from 2,542 participants (testing) | Apple validation paper, September 2024 |
| Wearables for OSA, AHI above 15 (pooled) | Sensitivity 0.87, specificity 0.86, AUC 0.93 (2,743 participants) | Sleep and Breathing meta-analysis, 2025 |
| Smartwatch algorithm, moderate-to-severe OSA | Sensitivity 92.3%, specificity 92.6%, accuracy 92.5% (90 adults) | Nature and Science of Sleep, 2025 |
| AASM position on consumer sleep technology | Cannot be used for diagnosis or treatment without FDA clearance and validation | AASM Position Statement, JCSM 2018 |
Context note: the Sleep and Breathing meta-analysis pooled 59 studies and includes medical-grade wearables, not only consumer watches (Sleep and Breathing, 2025). The Nature and Science of Sleep result came from a single Korean cohort with 53 usable datasets. The AASM statement dates from 2018 (most recent available position statement).
6. Adoption, Behavior Change and Orthosomnia
Accuracy limits have not slowed uptake. 48% of US adults have used a sleep tracking device such as a smartwatch or app, up from 35% in 2023, and 55% say they act on what the tracker tells them (AASM, Sleep tracking and sleepmaxxing survey, 2026). The 2023 wave found 77% of users considered trackers helpful and 68% changed behavior (AASM, 2023). When half the population acts on data whose epoch agreement kappa ranged from 0.21 to 0.65 in the healthy-adult studies above, the accuracy question becomes a public-health question.
The downside has a name. A 2024 Brain Sciences study of 523 adults, 35.8% of whom regularly used sleep trackers, estimated orthosomnia prevalence at 3.0%, 8.6% or 14.0% depending on the questionnaire cutoff (Brain Sciences, 2024). Ring demand shows the category’s momentum: Oura announced it surpassed 5.5 million rings sold in September 2025, up from 2.5 million in June 2024 (Oura Newsroom). Many users pair trackers with other sleep aids, as seen in our white noise statistics.
| Metric | Value | Source |
|---|---|---|
| US adults who have used a sleep tracking device | 48% (2025) vs 35% (2023) | AASM surveys, 2025 and 2023 |
| Adults acting on tracker data | 55% (2025); 68% changed behavior (2023) | AASM surveys, 2025 and 2023 |
| Users who found trackers helpful | 77% (2023) | AASM survey, 2023 |
| Tracker use by sex | Men 56%, women 40% | AASM survey, 2025 |
| Adults who lost sleep worrying about sleep problems | 76% | AASM survey, 2025 |
| Orthosomnia prevalence (523 adults) | 3.0%, 8.6% or 14.0% by cutoff | Brain Sciences, 2024 |
| US adults regularly wearing a smartwatch or fitness tracker | 21% (2019) | Pew Research Center, 2020 |
| Oura rings sold | More than 5.5 million (September 2025); 2.5 million (June 2024) | Oura Newsroom, 2024 and 2025 |
Context note: Pew’s figure comes from a June 2019 survey of 4,272 adults (most recent available data, Pew Research Center 2020), when 31% of adults in households earning 75,000 dollars or more wore one versus 12% under 30,000 dollars. Oura also reported a 5.2 billion dollar valuation with its December 2024 Series D and said in May 2026 it was on pace to pass five million paid members. The broader device market is covered in our wearables statistics.
Summary: Sleep Tracker Accuracy by the Numbers
| Metric | Value | Source |
|---|---|---|
| Sleep epoch sensitivity, six wrist devices | More than 90% | Sleep Advances, 2025 |
| Wake specificity, six wrist devices | 29.39% to 52.15% | Sleep Advances, 2025 |
| Epoch kappa range, six wrist devices | 0.21 to 0.53 | Sleep Advances, 2025 |
| Total sleep time overestimation range | 6.31 to 39.87 min | Sleep Advances, 2025 |
| Oura Ring Gen3 sleep-wake accuracy | 91.7% to 91.8% | Sleep Medicine, 2024 |
| Four-stage agreement, best device | Oura Gen3 76.3% (kappa 0.65) | Sensors, 2024 (Oura-funded) |
| Apple Watch S8 deep sleep bias | -43 min | Sensors, 2024 |
| Apple Watch S7 light sleep bias | -73.8 min | Sleep Medicine, 2026 |
| Wrist device meta-analysis | 24 studies, 798 participants | Journal of Clinical Sleep Medicine, 2025 |
| Oura pooled total sleep time difference | -2.97 min (not significant) | OTO Open, 2025 |
| Kappa in sleep clinic patients | Withings 0.12, Fitbit 0.07 | Journal of Sleep Research, 2025 |
| Fitbit kappa in insomnia | 0.00 | Journal of Sleep Research, 2025 |
| Device vs self-report concordance | 82.4% good sleepers vs 39.4% insomnia | Sleep Medicine, 2026 |
| Samsung Sleep Apnea Feature | FDA De Novo, February 6, 2024 | US FDA |
| Apple sleep apnea sensitivity / specificity | 66.3% / 98.5% | Apple validation paper, 2024 |
| Apple sensitivity, severe sleep apnea | 89.1% | Apple validation paper, 2024 |
| US adults who have used a sleep tracker | 48% (35% in 2023) | AASM survey, 2025 |
| Adults acting on tracker data | 55% | AASM survey, 2025 |
| Orthosomnia prevalence | 3.0% to 14.0% by cutoff | Brain Sciences, 2024 |
| Oura rings sold | More than 5.5 million | Oura Newsroom, 2025 |
Methodology and Sources
Every figure above was read during research for this article in a peer-reviewed paper (via PubMed Central or Europe PMC), a regulator database, a professional society release or a company publication. Several validation studies are industry-funded, and that is noted where relevant. Figures circulating on blogs without a primary document were excluded.
- Schyvens et al., Sleep Advances 2025: A performance validation of six commercial wrist-worn wearable sleep-tracking devices (funded by Flanders Innovation and Entrepreneurship, VLAIO)
- Robbins et al., Sensors 2024: Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults (funded by Oura Ring Inc.)
- Svensson et al., Sleep Medicine 2024: Validity and reliability of the Oura Ring Gen3 with OSSA 2.0
- Alhejaili et al., Sleep Medicine 2026: Performance of three consumer sleep-tracking devices compared with actigraphy and polysomnography
- Chinoy et al., Sleep 2021: Performance of seven consumer sleep-tracking devices compared with polysomnography
- Journal of Clinical Sleep Medicine 2025: Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis
- Agostinho et al., Behavioral Sleep Medicine 2026: Are wearable sleep-tracking devices reliable alternatives to polysomnography?
- OTO Open 2025: The Oura Ring versus medical-grade sleep studies
- Journal of Sleep Research 2025: Metrology of two wearable sleep trackers against polysomnography in patients with sleep complaints
- Sleep Medicine 2026: Concordance of wearable device sleep metrics with patient-reported sleep quality
- Nature and Science of Sleep 2025: Accuracy bias and factors influencing PSG and consumer device total sleep time; Journal of Clinical Sleep Medicine 2024, Fitbit Charge 4 in Huntington’s disease (DOI 10.5664/jcsm.11098); Nature and Science of Sleep 2025, smartwatch AHI estimation (DOI 10.2147/nss.s540460)
- Kainec et al., Sensors 2024: Evaluating accuracy in five commercial sleep-tracking devices
- Sleep and Breathing 2025: Diagnostic value of wearable and contactless technologies for detecting obstructive sleep apnea
- US FDA: De Novo DEN230041, Samsung Sleep Apnea Feature
- Apple: Sleep Apnea Notifications on Apple Watch validation paper (September 2024) and Apple Newsroom, September 2024
- American Academy of Sleep Medicine: Consumer Sleep Technology position statement (JCSM 2018), 2025 sleep tracking survey (published January 2026), 2023 sleep tracker survey
- Brain Sciences 2024: Prevalence of orthosomnia in a general population sample
- Pew Research Center: About one-in-five Americans use a smart watch or fitness tracker (2020)
- Oura: Newsroom press release list
- Data watch: Validation results are tied to specific device generations and firmware (for example Oura Gen3 with OSSA 2.0, Apple Watch Series 7 and 8, Fitbit Sense 2), so newer models and algorithm updates may perform differently. Studies disagree on Apple Watch staging: Robbins et al. found four-stage agreement of 75.0%, while Alhejaili et al. found weak agreement and a 73.8-minute light sleep underestimate in a different model and home setting. The Robbins study was funded by Oura and its lead author sits on Oura’s medical advisory board. The JCSM 2025 meta-analysis abstract reports signed mean differences without stating their direction; the 2026 Behavioral Sleep Medicine review states wearables overestimate total sleep time. The Chinoy (2021), AASM position statement (2018) and Pew (2019 fieldwork) figures are most recent available data. Apple’s FDA 510(k) record could not be read (the FDA database returned errors), so this article states only Apple’s own September 2024 announcement that marketing authorization was expected. Oura’s sales figures come from press release titles on its newsroom; the underlying Business Wire releases returned 403 errors. PubMed abstracts were read through the Europe PMC API because PubMed pages were blocked.
Last updated: October 3, 2026. We update this roundup quarterly, and the next refresh is expected when the next AASM sleep survey is released and new peer-reviewed validations of current-generation rings and watches appear in Sleep, Sleep Medicine and the Journal of Clinical Sleep Medicine.