Kygo Health

UPDATED 13 OCT 2026 · 6 BRANDS · 10 METRICS + FDA

Which wearable is actually accurate?

We read 78 validation studies and 14 FDA records on Oura, Apple Watch, Garmin, WHOOP, Fitbit and Samsung. No device wins everywhere, so compare the ones you're considering on the metrics you care about.

Research behind this page

Funding checked

validation studies

FDA records

metrics graded, plus FDA

Apple

21 studies

Garmin

20 studies

Fitbit

18 studies

Samsung

14 studies

Oura

10 studies

WHOOP

5 studies

Independent

Funded by the brand

Independent studies, no brand money

Brand-funded studies, flagged

Published in 2025 or 2026

115%

Worst calorie error in a 2026 lab test

Build your comparison

Compare wearables head-to-head.

Pick two brands and see every metric side by side. Open any metric for the full chart.

Best for Sleep stages, Sleep timing, Active HR, SpO2, Steps

None required

Best for Sleep stages, Sleep timing, HRV, Resting HR

$5.99/mo required

Overview

Sleep stages

Sleep timing

HRV

Resting HR

Active HR

SpO2

Steps

Calories

VO2 max

Skin temp

Breathing

FDA-cleared

FDA-cleared features

Tested against FDA 510(k) and De Novo decisions · checked Oct 2026

Apple has the most cleared features (5) and Oura has none, but cleared does not mean every case is caught.

Feature	Apple	Oura
ECG / AFib check	Series 4+
Irregular rhythm alerts
AFib History
Sleep apnea alerts	S9+, Ultra 2+, SE 3
Hypertension alerts	S9+, Ultra 2+
Blood pressure (wellness, not cleared)	Wellness Ring 5, no numbers
Cleared features	5	0

Apple

cleared

ECG / AFib check Series 4+
Irregular rhythm alerts
AFib History
Sleep apnea alerts S9+, Ultra 2+, SE 3
Hypertension alerts S9+, Ultra 2+

Oura

cleared

Blood pressure (wellness, not cleared) Ring 5, no numbers

From the FDA record (510(k) and De Novo decisions). Fitbit column counts Fitbit-branded devices; Google's Pixel Watch also has a cleared Loss of Pulse feature (3 in total). No wearable's blood oxygen feature is cleared.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Breathing rate

Tested against nasal airflow, chest bands or impedance · checked Oct 2026

Fitbit is the only brand with an independent test of the shipped feature (0.3 to 1 breath a minute high); WHOOP is second on a peer-reviewed company-funded test. No third qualifies.

Breathing rate

Overnight breathing rate, breaths per minute off

Independent test

Company-run test

Oura

Not tested

0.7/min

Breaths per minute off, shorter is better. Different tests and measures (average error, RMSE, bias), so not a ranking. Company tests are mostly healthy sleepers; independent tests include hospital patients. Samsung's figure is a research algorithm, not the watch feature. Garmin's readings tracked the reference poorly (correlation 0.28). Not charted: Apple Watch Series 7 gave no readings in two independent tests (Helmer 2024, AbiMansour 2024), and WHOOP has only a bias figure (Berryhill 2020, 1.8%).

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Skin temperature

Tested against research-grade iButton sensors · checked Oct 2026

No independent study has tested any brand's temperature sensor.

Wearable	Sensor test	Cycle tracking study	Grade
Apple	89% within 2 days Apple-funded, prototype
Oura	r² above 0.92 worn Own 2020 test	96.4% detected Oura staff	Company claim only

No independent study has tested any brand's temperature sensor. Cycle figures use different definitions and are not comparable across brands.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

VO2 max

Tested against lab metabolic cart (CPET) · checked Oct 2026

Garmin has the most validation at about 7% error. Apple reads low in all three independent studies, and Fitbit's Pixel Watch 3 came in at 8.7% in a Google-funded test.

Typical error

How far off a lab VO2 max test

Apple

13.2%

Not in this test:

Oura

Typical error against a lab treadmill test with gas analysis, shorter is better. Rows follow the medal order. Separate studies, so read each brand on its own. Medals: Garmin first on error (6.7% with a chest strap); Fitbit second on the Pixel Watch 3 (8.7%, Google-funded), which uses the outdoor-run method Google now ships, with the Charge 2 at 9.1% on the retired method; Apple third (13.2%, reads low in all three independent studies). Garmin: Engel 2026 (35 runners). Apple: Lambe 2026, Series 10 (40 adults), reading low like two earlier Apple studies. Fitbit: Klepin 2019 (60 people), on an algorithm Google retired in May 2026. Oura, WHOOP and Samsung have no independent VO2 max study.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Calories burned

Tested against indirect calorimetry (lab gas analysis) · checked Oct 2026

The least accurate number on any wearable. Fitbit is least bad, Samsung second, Apple third, and every brand is more than 10% off in every head-to-head test.

Workouts

Reads high or low

Daily totals

Workout calories, typical miss either way (the studies give no direction)

Cycling

Weight training

Reads low

Reads high

Apple

±36%

±112%

Not in this test:

Oura

Average over or under the real burn, same 62 people

Endurance

Weight training

Reads low

Reads high

Apple

+14%

+112%

Not in this test:

Oura

Calories across whole days of normal life

Whole-day burn

Activity calories only

Reads low

Reads high

Oura

±13%

Not tested

Apple

Not tested

+26%

Typical miss either way (striped) and average over or under (solid). Workouts: cycling with the same 58 people on all four brands (Kostrna 2026) and weight training, 70 men (Yun 2026). Reads high or low: 62 men wore all four through endurance and weight sessions (Lee 2026; its abstract says "underestimated" but its tables show all four reading high). Daily: each brand used a different reference, so do not compare them.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Step counting

Tested against hand count (lab) or research pedometer (daily life) · checked Oct 2026

Garmin counts steps most accurately: 3% over in daily life across 59 studies and 1 to 2% in the lab. Apple is second (7%), Samsung third. Fitbit and especially Oura overcount in daily life.

Daily life

Lab walking

Steps over whole days of normal life

Average over or under

Typical miss

Reads low

Reads high

Apple

+2% · ±6%

Oura

+19% · ±50%

Against a hand or video count

Typical miss

Reads low

Reads high

Apple

±9.3% slow walk

Not in this test:

Oura

Separate studies, so read each brand on its own. Medals follow the pooled free-living result across 59 studies (Paluch 2026: Garmin +3%, Apple +7%, Fitbit +15%), with Samsung third on its single 24-hour study. Daily life: Miwa 2026, Hong 2024 (Apple and Samsung in the same 24-hour study: 6.4% vs 10.5%), Kristiansson 2023; Oura figures predate its 2025 step update. Garmin: Vivofit, a 2014 model, at home (An 2017). Lab: Fitbit and Garmin on normal treadmill walks (Cheung 2025, de Leon 2026); Apple on slow and shuffle walks (Rowe 2025), which is harder.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

SpO2

Tested against arterial blood or hospital oximeter · checked Oct 2026

Apple is the most accurate (2.2 points off in a low-oxygen test), Samsung second (3.8, reads low) and Garmin a distant third. Every watch is least reliable when oxygen is truly low.

Typical error

Reads high or low

How far off a medical reference

Apple

2.2%

Oura

Not tested

How far blood oxygen reads high or low on average

Reads low

Reads high

Apple

-0.1 pts

Oura

Not tested

Typical error: oxygen points off, shorter is better; Apple and Samsung come from the same low-oxygen lab test (Walzel 2023), which sets the medals, and Garmin (clinic patients at rest, Jiang 2023) is marked as a different study. Reads high or low: the average gap in oxygen points; Fitbit (Hermans 2024, COPD patients overnight) and Garmin are separate studies, so compare direction more than size. Oura and WHOOP have no independent SpO2 study; Oura has only its own unpublished comparison.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Heart rate during activity

Tested against chest strap or 12-lead ECG · checked Oct 2026

Apple edges Garmin and Fitbit across three head-to-head tests, and the top four brands sit within 2 to 7% of each other. Model matters: the best and worst devices in a ten-device test were both Fitbits.

Daytime

By sport

Heart rate during daytime activity

Apple

7.3%

Oura

15.0%

Same watch, same people, real training

Running

1.2%

Cycling

8.1%

Tennis

8.9%

Orienteering

9.5%

Badminton

16.2%

Soccer

17.5%

Typical error, shorter is better. Fitbit, Garmin, Apple and Oura: one ten-device test against a chest strap (Gielen 2026), read as tiers: Fitbit and Garmin Tier 1, Apple Tier 2, Oura Tier 4. Medals combine this test with two other head-to-heads on the same people: cycling against a chest strap (Kostrna 2026: Apple 2.5 bpm, Garmin 2.9, Samsung 3.4, Fitbit 3.9) and a treadmill against a 12-lead ECG (Van Oost 2025), which puts Apple first, Garmin second, Fitbit third; the four are within 2 to 7% of each other. WHOOP, marked as a different study: Van Oost 2025. Samsung is not charted: its own test was heart patients in a hospital stress test (Kim 2023, under 2%), and it was 3rd of 4 cycling and worst of 4 on a treadmill (Jeong 2026). By sport: one Fitbit in real training (Ceugniez 2025).

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Resting heart rate

Tested against chest-strap or lead II ECG · checked Oct 2026

Oura is the most accurate overnight (1.1 bpm, 1.9% error), WHOOP second (1.8 bpm) and Apple third; Garmin was 5.4 bpm off in its one lab night.

Typical error

Reads high or low

Resting heart rate overnight, how far off an ECG

Oura

1.1 bpm

Apple

1.5 bpm

different study

How far resting heart rate reads high or low on average

Reads low

Reads high

Oura

-0.9 bpm

Apple

+0.5 bpm

different study

Beats per minute off an ECG overnight; shorter is better. Oura (138 nights) and WHOOP (288 nights): one independent study (Dial 2025), which sets the medals. Apple and Garmin, marked as a different study: one night in a lab that receives WHOOP funding (Miller 2022). Fitbit (Charge 4, 10 people with Huntington's disease in a lab sleep study, Doheny 2024) and Samsung (Gear Sport asleep, research software, Sarhaddi 2022) measured heart rate through the night rather than one resting value, also marked. Reads high or low: the average gap in bpm, for all six brands; Fitbit (Hermans 2024), Samsung (Sarhaddi 2022, research software) and Apple and Garmin (Miller 2022) are separate studies, so compare direction more than size. Older Fitbit studies: Charge 2 0.1 to 0.9 bpm off overnight (Haghayegh 2019, Stucky 2021), Charge HR 0.7 bpm low over a day (Benedetti 2021).

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Overnight HRV

Tested against Polar H10 chest-strap ECG · checked Oct 2026

Oura Ring 4 matched a chest-strap ECG almost perfectly over 139 nights (6% error). WHOOP is second (8%) and Garmin third (10.5%) in the same independent test; Apple, Fitbit and Samsung have no comparable figure.

Agreement

Reads high or low

Overnight HRV, agreement with a chest-strap ECG

Oura

0.99

Apple

0.67

different study

How far HRV reads high or low on average

Reads low

Reads high

Oura

-1.0 ms

Apple

-8.3 ms

different study

Agreement: 1 is perfect; one independent study (Dial 2025, 13 people, 536 nights) sets the medals. Fitbit (Hermans 2024) and Apple (Miller 2022) are marked as different studies that report the closely related ICC; Samsung has no agreement figure. Reads high or low: the average gap from a chest strap or ECG in milliseconds. Oura, WHOOP and Garmin come from Dial 2025; Apple (O'Grady 2024, morning readings), Fitbit (Hermans 2024, healthy adults) and Samsung (Sarhaddi 2022, research software, not the watch feature) are separate studies, so compare direction more than size. For scale: Oura read 1 to 2.5 ms low in Dial but an older Oura (Gen 2) read 10 ms low in a one-night lab study.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Sleep timing

Tested against polysomnography (lab sleep study) · checked Oct 2026

Oura knows when you were asleep best: total sleep within 2 minutes of a lab study. Fitbit is second, Apple third because it misses half of the time you lie awake.

Awake time caught

Total sleep time

Sleep efficiency

Time to fall asleep

Awake at night

Asleep vs awake

Time awake in bed correctly caught (higher is better)

Healthy adults

Sleep problems

Oura

73.0%

46.2%

Apple

52.4%

52.2%

Total sleep time vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-2 min

+12 min

Apple

+8 min

+20 min

Sleep efficiency vs a lab study, in percentage points (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-1.1 pts

+3.0 pts

Apple

+2 pts

+4.4 pts

Time to fall asleep vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

0 min

-3 min

Apple

+2 min

Time awake during the night vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-2 min

-9 min

Apple

-10 min

-21 min

Asleep or awake, called right (agreement with a lab study)

Healthy adults

Apple

93%

Oura

89%

Device minus the lab study. Healthy adults: Svensson 2024 (Oura, Oura-funded, 96 adults over several nights), Robbins 2024 (Apple, Fitbit; Oura-funded), Miller 2022 (WHOOP 3.0, Garmin Forerunner 245, Apple asleep vs awake; lab receives WHOOP support), Chinoy 2021 (Garmin Vivosmart 3 efficiency, fall asleep, awake at night), Miller 2021 (WHOOP 2.0 fall asleep) and Stone 2020 (WHOOP 2.0 efficiency, vs an EEG headband rather than a lab study). Sleep problems: Schyvens 2025 (Apple, Fitbit Charge 5, WHOOP 4.0), Herberger 2025 (Oura), Lee 2023 (Samsung efficiency), Byun 2023 (Samsung total sleep) and Budig 2022 (Garmin Forerunner 945). Asleep vs awake runs high for every device because most of the night is sleep.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Sleep stages

Tested against polysomnography (lab sleep study) · checked Oct 2026

Oura stages sleep most accurately in healthy sleepers, Apple is second and Fitbit third. No tracker matches a lab, and every one gets worse in people with sleep problems.

Overall staging

Deep sleep

REM sleep

Match with a lab sleep study (kappa, 1 is perfect)

Healthy adults

Sleep problems

Two human scorers: 0.75

Oura

0.65

0.35

Apple

0.60

0.53

Deep sleep correctly caught

Healthy adults

Sleep problems

Oura

64.0%

50.0%

Apple

50.5%

50.7%

REM sleep correctly caught

Healthy adults

Sleep problems

Oura

72.6%

55.0%

Apple

82.6%

68.6%

Healthy adults: Robbins 2024 (Oura-funded, one lab test of Oura, Apple and Fitbit), Svensson 2024 (Oura-funded; Oura deep and REM) and Miller 2022 (WHOOP, Garmin; lab receives WHOOP support). Sleep problems: Schyvens 2025 (Apple, Fitbit, WHOOP, Garmin; about half with sleep apnea), Lee 2023 (Oura kappa, Samsung) and Herberger 2025 (Oura deep and REM, sleep-clinic patients). The independent Topalidis 2026 preprint orders Oura 0.59, WHOOP 0.54, Garmin 0.41, consistent with the healthy column.

Same figures as the full section, filtered to your brands. A dashed circle means no independent study.

Full section and studies

Free 7-day trial

Every metric, side by side

Peer-reviewed results, funding flagged in each section · green marks the better one

Metric

Apple

Oura

Sleep stages

κ 0.60 healthy, best REM (83%); 0.53 clinical

κ 0.65 healthy; 0.35 clinical

Sleep timing

Total sleep +8 min; catches 52% of wake

Total sleep -2 min; catches 73% of wake

HRV

29% error, different statistic (SDNN, mornings)

6% error, CCC 0.99 (Ring 4)

Resting HR

5.9%, 3.7 bpm off, mornings lying down (Series 9)

1.9%, 1.1 bpm off overnight (Ring 4)

Active HR

2.5 bpm cycling (best of 4); 7.3% ten-device test

15% ten-device test (Gen 3)

SpO2

Arms 2.2, bias -0.1 (Series 8)

Company claim only

Steps

+7% daily life pooled; 6.4% in 36 adults

50 to 69% off, pre-2025 algorithm

Calories

26% treadmill, 36% cycling, 112% weights

No calorimetry test

VO2 max

13.2%, reads low (Series 10)

Company claim only

Skin temp

No sensor test

Company claim only

Breathing

No readings (2 hospital studies)

Company claim only

FDA-cleared

5 cleared

None

Figures summarise each metric's section. Tap a metric name for the full chart.

Free 7-day trial

No device wins everywhere. Kygo reads them all.

Whichever wearable you own, Kygo connects to it and cross-checks your key metrics against what you actually eat, train and sleep.

Works with

Oura

Apple

Fitbit

Garmin

WHOOP

Polar

Google

Health

The short answer

Each brand wins something, none wins everything.

How each brand graded on every metric, from peer-reviewed studies with funding flagged. Every device is weak on calories, and the newest models are mostly untested.

Metric

Apple

Oura

Garmin

WHOOP

Fitbit

Samsung

Sleep stages

Sleep timing

HRV

Resting HR

Active HR

SpO2

Steps

Calories

VO2 max

Skin temp

Breathing

FDA-cleared

Buy

Amazon

Top three

Good

Fair

Weak

Not tested, or no qualifying test

One podium per metric, ranked on the numbers across every measure of that metric (sleep is two: telling stages apart, and knowing when you were asleep). Brand-funded studies count and are flagged in each section; research algorithms and unpublished company tests do not. Calories medals mean least bad: every brand is more than 10% off. Breathing rate has only two qualifying brands, and no brand has an independent skin temperature test.

Each brand in detail

Strengths, gaps, subscription and studies

Apple Watch

The most accurate for workout heart rate and blood oxygen, second for steps and sleep stages, with the most FDA-cleared features (5).

Where it leads

Most accurate workout heart rate: lowest error of four watches cycling (2.5 bpm, Series 8) and near-zero pooled bias across 22 studies (Lambe 2026)
Most accurate blood oxygen: Arms 2.2 in a low-oxygen test (Series 8) and 2.4 against arterial blood in 112 patients
Second for sleep stages (κ 0.60) with the best REM detection of any brand (83%), healthy adults
Second for steps: +7% in daily life pooled across 59 studies, 6.4% in a 36-adult study
5 FDA-cleared features: ECG, irregular rhythm alerts, AFib History, sleep apnea and hypertension notifications (most of any brand)

Where it falls short

Third for sleep timing: misses about half of the time you lie awake (52%) and reads total sleep 8 min high
VO2 max underestimated in all 3 independent studies (MAPE 13 to 16%), third behind Garmin and Fitbit
Weight-training calories off by 112% on average (Series 8, Yun 2026)
Clinical-sample sleep staging drops to κ=0.30 (Lee 2023)
Over-read 85% of blood oxygen readings when true oxygen was below 88% (Series 7)
Cleared alerts miss many cases: hypertension 41% sensitivity, sleep apnea 66% (FDA summaries)
Hypertension notifications on Series 12 / Ultra 4 on hold in 33 countries (not the US)
Series 10, 11 and Ultra 3 have no heart rate or sleep validation

US watches have computed blood oxygen on the paired iPhone since August 2025. Every Apple SpO2 study predates that change or was run outside the US.

Subscription None required · Studies Schyvens 2025, Lee 2023, Robbins 2024, Gielen 2026, Lambe 2026, Miller 2022, Hong 2024, Miwa 2026, Rowe 2025, Kostrna 2026, Yun 2026, Lee 2026, Perez 2025, Apple 2024, Caserman 2024, Lambe 2025, Lambe 2026, Walzel 2023, Jiang 2023, Jiang 2026, Wang 2025, O'Grady 2024

Oura Ring

The most accurate for everything measured overnight: sleep stages, sleep timing, HRV and resting heart rate. Nothing to show in the daytime.

Where it leads

Most accurate overnight HRV: 6% error, agreement 0.99 (Ring 4, 139 nights, independent)
Most accurate resting heart rate overnight: 1.1 bpm, 1.9% (Ring 4; WHOOP 1.8 bpm in the same study)
Most accurate sleep stages in healthy sleepers (κ 0.65 vs Apple 0.60, Fitbit 0.55 in Oura's own study; independent preprint agrees on the order)
Most accurate sleep timing: total sleep within 2 min, catches 73% of wake (96 adults, multi-night)
Only brand with a published temperature sensor test (its own)
Lightest Oura yet: Ring 5 weighs 2 to 2.6 g (Oura spec)

Where it falls short

Steps 50 to 69% off before its May 2025 algorithm change; nothing published since
Daytime heart rate error 15% (Gen 3, Gielen 2026), bottom tier
Sleep staging κ=0.35 in a clinical sample; 53% accuracy in sleep-clinic patients
Calorie algorithm changed May 2025 and has not been independently tested
No independent blood oxygen or respiratory rate validation
No FDA-cleared features (Oura IPO filing, Sept 2026); Ring 5 blood pressure is wellness only
Ring 5 has zero validation of any kind

Sleep κ=0.65 is from an Oura-funded study of screened healthy sleepers that used the same method as the independent studies. In people with sleep complaints Oura scored κ=0.35 (Lee 2023) and 53% accuracy in sleep-clinic patients (Herberger 2025).

Subscription $5.99/mo required · Studies Robbins 2024, Lee 2023, Herberger 2025, Khan 2025, Dial 2025, Miller 2022, Gielen 2026, Kristiansson 2023, Oura 2020, Thigpen 2025, Oura 2020, Niela-Vilén 2022

Garmin

The most accurate for steps and VO2 max (with a chest strap), second for workout heart rate.

Where it leads

Most accurate steps: +3% in daily life pooled across 59 studies, 1 to 2% on a treadmill in 258 adults
Most accurate VO2 max with a chest strap: 6.7% error in 35 runners, about 7% in two more lab tests
Second for workout heart rate: 2.9 bpm cycling, 6.3% in a ten-device test (Vivoactive 5)
Third for overnight HRV (10.5% error, CCC 0.87, Fenix 6)
FDA-cleared ECG app on select watches (2023)
Up to 11 days of battery (Vivoactive 5, Garmin spec)

Where it falls short

Weakest sleep: stages κ 0.21 to 0.25, catches 27% of wake, reads total sleep 44 min high
Wrist-only VO2 max read 13 mL/kg/min high in 207 mostly older women (56% error)
Largest workout calorie error in a four-brand test (77.6% cycling) and 115% in weight training
Blood oxygen read low at rest (Venu 2s) and high at altitude (Fenix 5X)
Respiratory rate barely tracked real breathing in hospital patients (r 0.28)
11.5% of ECGs came back inconclusive in the FDA study
No flagship since the Fenix 6 has peer-reviewed validation

Every Garmin figure here is Fenix 5X/6, Forerunner 245, Vivosmart 4/5, Vivoactive 4/5 or Venu 2s. Fenix 8, Forerunner 970 and Venu 4 have nothing beyond one 2026 hospital preprint.

Subscription None required · Studies Schyvens 2025, Dial 2025, Miller 2022, Gielen 2026, de Leon 2026, Rider 2025, Kostrna 2026, Yun 2026, Lee 2026, Perez 2025, Carrier 2025, Carrier 2023, Engel 2026, Jiang 2023, Schiefer 2021, Knevels 2026, Helmer 2024, Svarre 2020, An 2017, Montes 2020

WHOOP

Second for HRV, resting heart rate and breathing rate. No peer-reviewed test for anything in the daytime.

Where it leads

Second for overnight HRV: 8% error, agreement 0.94 (4.0, 289 nights, independent)
Second for resting heart rate overnight: 1.8 bpm, 3.0% (same study)
Second for breathing rate: +0.1 breaths/min vs sleep-lab belts (Strap 2.0, WHOOP-funded, peer reviewed)
Caught 70% of deep sleep in a mixed sample (Schyvens 2025)
FDA-cleared ECG, WHOOP MG only
Continuous 24/7 wear, no screen

Where it falls short

Daytime heart rate MAPE 8.5% vs 12-lead ECG, second worst of five (Van Oost 2025)
Total-sleep-time bias +24.5 min (Schyvens 2025)
No published calorie, step, SpO2 or independent VO2 max validation
Blood pressure is a wellness estimate, not cleared; FDA dropped its warning letter in June 2026 after WHOOP changed the feature
WHOOP 5.0 and MG have no validation; VO2 max is 5.0/MG only

The widely quoted "99.7% heart rate accuracy" comes from one night of sleep on WHOOP 3.0 (Miller 2022), in a lab that receives WHOOP support.

Subscription From $199/yr · Studies Schyvens 2025, Dial 2025, Miller 2022, Van Oost 2025, Berryhill 2020

Fitbit

The least bad for calories and the only brand with an independent breathing rate test. Second for sleep timing and VO2 max.

Where it leads

Least bad calories: wins three of five head-to-head comparisons, weight-training reads only 4% high
Only brand with an independent breathing rate test: 0.3 to 1 breath/min high (Charge 4)
Second for sleep timing (total sleep -3 min, catches 68% of wake) and third for stages (κ 0.55)
Second for VO2 max: 8.7% (Pixel Watch 3, Google-funded), 9.1% in 60 adults (Charge 2)
Third for workout heart rate; the Charge 6 was best of ten devices (5.5%) but the Sense 2 last of four cycling
Daily calories 9 to 13% off doubly labelled water, the gold standard (Alta HR, Shook 2022)
FDA-cleared ECG (Sense, Sense 2, Charge 5 and 6) and irregular rhythm alerts on most models

Where it falls short

Worst device in the same ten-device heart rate test was also a Fitbit (Inspire 3, 16.5%)
Overcounts steps by 15% in free living pooled across 59 studies, up to 26% in single studies
Smallest average calorie bias, but large misses on single workouts and implausible readings in one study
Pooled heart rate under-reads by 3.4 bpm with wide limits (Chevance 2022)
Blood oxygen over-reads by about 2 points in COPD (Charge 4, ICC 0.32)
No sleep or step validation for any model newer than the Sense / Inspire 2

Since 19 May 2026 Google derives Fitbit VO2 max only from outdoor GPS runs. That method has no published validation.

Subscription Google Health Premium optional ($9.99/mo) · Studies Schyvens 2025, Lee 2023, Robbins 2024, Gielen 2026, Chevance 2022, Cheung 2025, Miwa 2026, Kostrna 2026, Yun 2026, Lee 2026, Perez 2025, Shook 2022, Lambe 2026, Freeberg 2019, Hermans 2024, Natarajan 2021, Nasr 2025, Doheny 2024

Samsung Galaxy Watch

Second for blood oxygen and calories, third for steps. No test of the shipped feature for HRV, resting heart rate, VO2 max or breathing rate.

Where it leads

Second for blood oxygen: Arms 3.8 in a low-oxygen test (Galaxy Watch 5), 1.8 vs arterial blood under the FDA protocol (Samsung-funded)
Second for calories: most accurate on steady running and walking (9% max treadmill, 0.5% walking)
Third for steps: 10.5% in daily life (Galaxy Watch 4)
Near-ECG heart rate in heart patients on a stress test (Galaxy Watch 4, ICC 0.997), but 3rd of 4 cycling and worst of 4 on a treadmill (Galaxy Watch 6)
FDA-cleared ECG, irregular rhythm alerts and sleep apnea risk check (first sleep apnea feature of its kind, De Novo 2024; v2.0 cleared July 2026 for Galaxy Ring too)

Where it falls short

Calories roughly double true burn on weights (85%) and miss by 67% cycling
Heart rate agreement drops above 160 bpm
Blood oxygen read low in an independent hypoxia test (Galaxy Watch 5, bias -2.6)
Respiratory rate validation was Samsung-funded and tested a research algorithm, not the consumer feature
US blood pressure is a wellness feature, not FDA-cleared, and needs cuff calibration every 28 days
No independent VO2 max, step or sleep validation for any current model; Galaxy Watch 8 and Galaxy Ring have nothing

Subscription None required · Studies Lee 2023, Kim 2023, Kostrna 2026, Yun 2026, Lee 2026, Perez 2025, Ferreira 2026, Walzel 2023, Browne 2025, Browne 2024, Jung 2023, Kim 2026, Sarhaddi 2022, Hong 2024

The Amazon links are affiliate links. As an Amazon Associate, Kygo Health earns from qualifying purchases, at no extra cost to you.

True for every device

What no brand can fix.

Six findings that hold across every wearable here. Read these before you trust any single number.

Population

Who wears it beats which brand

Universal

Calories are the weakest number

Funding

Who paid for the study matters

Coverage gap

The newest models are unvalidated

Blood oxygen

SpO2 is least reliable when it matters

Regulation

Cleared does not mean it catches everything

Population

0.65 → 0.35

Oura's sleep-staging score in healthy sleepers, then in people with sleep complaints

Who wears it beats which brand

The same device scores far lower in people with sleep complaints, across studies: Oura falls from κ 0.65 to 0.35, Apple from 0.53 to 0.30.

Robbins 2024 · Lee 2023 · Schyvens 2025

Universal

115%

The worst workout calorie error in a 2026 lab test

Calories are the weakest number

In 2026 lab tests weight-training calories were off by up to 115%, and Apple, Samsung and Garmin read high in cycling. Fitbit was close only on average. Use calories as a trend only.

Kostrna 2026 · Yun 2026

Funding

Sources here funded or written by the brand being tested

Who paid for the study matters

15 sources here were funded or written by the brand being tested: 11 studies and 4 company documents. Each one is flagged.

Funding checked per source

Coverage gap

2 to 4 yrs

How far peer review lags new hardware

The newest models are unvalidated

Oura Ring 5, WHOOP 5.0, Apple Watch Series 10 to 12 and Galaxy Watch 8 have little or no independent data. Peer review lags hardware by 2 to 4 years.

Validation coverage, Sept 2026

Blood oxygen

85%

Of readings below 88% oxygen, the share an Apple Watch Series 7 over-read (Jiang 2026)

SpO2 is least reliable when it matters

An Apple Watch over-read most readings when oxygen was truly low, and no watch has proven reliable in that range, so a normal reading is not reassurance if you feel unwell.

Jiang 2023 · Jiang 2026

Regulation

41%

Of people with hypertension, the share Apple's alert caught in its own FDA study

Cleared does not mean it catches everything

Apple's hypertension alert caught 41% of cases in its own FDA study, its sleep apnea alert 66%. No alert is not an all-clear.

FDA K250507 · K240929

1 of 6

The metrics

2 of 6

3 of 6

4 of 6

5 of 6

6 of 6

Ten metrics plus FDA, one section each.

The verdict, every brand against the lab gold standard, and who funded each study. Jump to the one you care about.

Sleep stages

Sleep timing

HRV

Resting HR

Active HR

SpO2

Steps

Calories

VO2 max

Skin temp

Breathing

FDA-cleared

Metric 01 · Sleep stages

Sleep stages: Oura leads; who you are matters more.

Oura stages sleep most accurately in healthy sleepers, Apple is second and Fitbit third. No tracker matches a lab, and every one gets worse in people with sleep problems.

Checked Oct 2026 · Tested against polysomnography (lab sleep study)

Overall staging

Deep sleep

REM sleep

Match with a lab sleep study (kappa, 1 is perfect)

Healthy adults

Sleep problems

Two human scorers: 0.75

Oura

0.65

0.35

Apple

0.60

0.53

Fitbit

0.55

0.41

WHOOP

0.44

0.37

Samsung

Not tested

0.42

Garmin

0.25

0.21

Deep sleep correctly caught

Healthy adults

Sleep problems

Oura

64.0%

50.0%

Apple

50.5%

50.7%

Fitbit

61.7%

51.5%

WHOOP

62.0%

69.6%

Samsung

Not tested

47.5%

Garmin

28.0%

47.5%

REM sleep correctly caught

Healthy adults

Sleep problems

Oura

72.6%

55.0%

Apple

82.6%

68.6%

Fitbit

67.6%

60.0%

WHOOP

66.0%

62.0%

Samsung

Not tested

62.7%

Garmin

50.0%

33.1%

What this means for you

Treat the stage breakdown as a rough guide, especially if you sleep poorly. Oura is the most accurate stager in healthy sleepers; Apple catches REM best but scores half of deep sleep as light.

Sleep Tracker Accuracy

Dedicated tool · 7 devices · head-to-head builder

Open

More data

The full picture, secondary tests and notes

No tracker stages sleep like a lab. In the one study that put Oura, Apple and Fitbit on the same sleepers, Oura scored highest (κ 0.65), Apple second (0.60, with the best REM detection) and Fitbit third (0.55); WHOOP, Garmin and Polar are well behind. Every device drops sharply in people with sleep problems, where Apple holds up best.

One podium: Oura first (κ 0.65), Apple second (0.60, best REM detection at 83%, but half of deep sleep scored as light), Fitbit third (0.55), all from one Oura-funded lab study of healthy sleepers that used the same method as the independent studies; the independent Topalidis 2026 preprint orders Oura, WHOOP and Garmin the same way. WHOOP (0.44), Garmin (0.25) and Polar (0.28) are well behind. Every device drops in people with sleep problems: Apple holds up best there (0.53), Oura falls to 0.35.

Studies behind this

7 sources, funding labelled

Oura-funded

Accuracy of three commercial wearable devices for sleep tracking in healthy adults

Robbins R, et al. · 2024 · Sensors · n=35

Oura-funded

Validity of the Oura Ring Gen 3 against polysomnography over multiple nights in healthy adults

Svensson T, et al. · 2024 · Sleep Medicine · n=96

Independent

Performance of six consumer sleep trackers vs polysomnography

Schyvens AM, et al. · 2025 · Sleep Advances · n=62

Independent

Accuracy of 11 wearable, nearable and airable consumer sleep trackers

Lee T, et al. · 2023 · JMIR mHealth and uHealth · n=75 (clinical, mean AHI 18)

Independent

Oura Ring Gen 3 in sleep-clinic patients

Herberger S, et al. · 2025 · Scientific Reports · n=45

Independent

Six wearables vs polysomnography and ECG for sleep, heart rate and HRV

Miller, et al. · 2022 · Sensors · n=53

Independent

Head-to-head comparison of eight consumer sleep trackers against polysomnography

Topalidis P, et al. · 2026 · Preprint · n=18

Metric 02 · Sleep timing

Sleep timing: Oura, then Fitbit, then Apple.

Oura knows when you were asleep best: total sleep within 2 minutes of a lab study. Fitbit is second, Apple third because it misses half of the time you lie awake.

Checked Oct 2026 · Tested against polysomnography (lab sleep study)

Awake time caught

Total sleep time

Sleep efficiency

Time to fall asleep

Awake at night

Asleep vs awake

Time awake in bed correctly caught (higher is better)

Healthy adults

Sleep problems

Oura

73.0%

46.2%

Fitbit

67.7%

47.5%

Apple

52.4%

52.2%

WHOOP

56.0%

40.1%

Samsung

Not tested

48.1%

Garmin

27.0%

29.4%

Total sleep time vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-2 min

+12 min

Fitbit

-3 min

+11 min

Apple

+8 min

+20 min

WHOOP

-12 min

+25 min

Samsung

Not tested

-25 min

Garmin

+44 min

+83 min

Sleep efficiency vs a lab study, in percentage points (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-1.1 pts

+3.0 pts

Fitbit

-1 pt

+3.4 pts

Apple

+2 pts

+4.4 pts

WHOOP

+7.4 pts

+4.1 pts

Samsung

Not tested

-0.4 pts

Garmin

+10.1 pts

+14.9 pts

Time to fall asleep vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

0 min

-3 min

Fitbit

0 min

+7 min

Apple

+2 min

WHOOP

0 min

-11 min

Garmin

-1 min

-10 min

Not in this test:

Samsung

Time awake during the night vs a lab study, in minutes (closer to 0 is better)

Healthy adults

Sleep problems

Reads low

Reads high

Oura

-2 min

-9 min

Fitbit

+4 min

-17 min

Apple

-10 min

-21 min

WHOOP

Not tested

-19 min

Garmin

-48 min

-50 min

Not in this test:

Samsung

Asleep or awake, called right (agreement with a lab study)

Healthy adults

Apple

93%

Fitbit

91%

Oura

89%

Garmin

89%

WHOOP

86%

Not in this test:

Samsung

What this means for you

Total sleep time and sleep efficiency are the dependable numbers on Oura, Fitbit and Apple. On any brand, the minutes awake figure is the one to doubt: every device misses a third to two thirds of the time you lie awake.

More data

The full picture, secondary tests and notes

Telling asleep from awake is the part of sleep tracking that drives your sleep score. Oura leads (total sleep within 2 minutes, 73% of wake caught), Fitbit is second and Apple third: Apple reads total sleep 8 minutes high and misses about half of wake. WHOOP and Garmin overestimate sleep by 12 and 44 minutes. In people with sleep problems every brand reads sleep high and catches half of wake at best.

One podium: Oura first (total sleep within 2 minutes, catches 73% of wake), Fitbit second (-3 minutes, 68% of wake), Apple third (+8 minutes; misses about half of wake, so it reads time awake at night 10 minutes low). Fitbit beats Apple on five of six timing measures; Apple's one win, asleep vs awake agreement, runs high for every device because most of the night is sleep. WHOOP (-12 minutes, 56% of wake) and Garmin (+44 minutes, 27% of wake) are well behind. In people with sleep problems every brand reads total sleep 11 to 25 minutes high (Garmin 83), and no brand catches more than 52% of wake.

Studies behind this

12 sources, funding labelled

Oura-funded

Validity of the Oura Ring Gen 3 against polysomnography over multiple nights in healthy adults

Svensson T, et al. · 2024 · Sleep Medicine · n=96

Oura-funded

Accuracy of three commercial wearable devices for sleep tracking in healthy adults

Robbins R, et al. · 2024 · Sensors · n=35

Independent

Six wearables vs polysomnography and ECG for sleep, heart rate and HRV

Miller, et al. · 2022 · Sensors · n=53

Independent

Performance of seven consumer sleep-tracking devices compared with polysomnography

Chinoy ED, et al. · 2021 · Sleep · n=34

Independent

A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults

Miller DJ, et al. · 2021 · Biosensors · n=6

Independent

Evaluations of commercial sleep technologies for objective monitoring during routine sleeping conditions

Stone JD, et al. · 2020 · Nature and Science of Sleep · n=5

Independent

Performance of six consumer sleep trackers vs polysomnography

Schyvens AM, et al. · 2025 · Sleep Advances · n=62

Independent

Oura Ring Gen 3 in sleep-clinic patients

Herberger S, et al. · 2025 · Scientific Reports · n=45

Independent

Accuracy of 11 wearable, nearable and airable consumer sleep trackers

Lee T, et al. · 2023 · JMIR mHealth and uHealth · n=75 (clinical, mean AHI 18)

Independent

Validity of the Galaxy Watch 2 for sleep in patients with obstructive sleep apnea

Byun JH, et al. · 2023 · Chronobiology International · n=127

Independent

Validity of two consumer multisport activity tracker and one accelerometer against polysomnography

Budig M, et al. · 2022 · Sensors · n=49

Independent

Oura Ring sleep measurement: systematic review and meta-analysis

Khan S, et al. · 2025 · OTO Open · n=6 studies, 388 people

Metric 03 · HRV

Overnight HRV: Oura leads.

Checked Oct 2026 · Tested against Polar H10 chest-strap ECG

Agreement

Reads high or low

Overnight HRV, agreement with a chest-strap ECG

Oura

0.99

WHOOP

0.94

Garmin

0.87

Fitbit

0.69

different study

Apple

0.67

different study

Not in this test:

Samsung

How far HRV reads high or low on average

Reads low

Reads high

WHOOP

-0.8 ms

Oura

-1.0 ms

Garmin

-1.8 ms

Fitbit

-7 ms

different study

Apple

-8.3 ms

different study

Samsung

-18.2 ms

different study

What this means for you

These are sleep-time numbers, so use overnight HRV only. Most watches and rings read increasingly low when your HRV is high, so very fit people should follow the trend rather than single nights.

More data

The full picture, secondary tests and notes

Oura Ring 4 matched a chest-strap ECG almost perfectly overnight (6% error, agreement 0.99). WHOOP was second (8%, 0.94) and Garmin third (10.5%, 0.87) in the same 536-night independent study; Polar was fourth. Apple reports a different statistic (SDNN, mornings) and was 29% off, Fitbit has no current-model error figure and Samsung has no test of the watch feature.

Oura Ring 4 is the clear leader for overnight HRV. Most devices read increasingly low at high HRV (Apple, Oura, Polar and Garmin in Miller 2022; WHOOP, which supplied raw beat data, did not), which matters for very fit people. These are sleep-time numbers only.

Oura Ring 4: CCC 0.99, 5.96% error (138 nights); Gen 3: CCC 0.97, 7.15% (470 nights), Dial 2025. Ring 5 has not been tested; the Ring 4 figures do not transfer to a redesigned sensor.
WHOOP 4.0: CCC 0.94, 8.17% (Dial 2025). WHOOP 3.0 scored ICC 0.99 in a lab that receives WHOOP funding (Miller 2022); its "99% HRV" claim traces to that study. WHOOP 5.0 and MG are untested.
Garmin Fenix 6: CCC 0.87, 10.52% (Dial 2025). Forerunner 245: ICC 0.24 in Miller 2022, but that was a 3-minute reading while awake before bed, not overnight.
Apple Apple reports SDNN while the others report RMSSD, so Apple HRV numbers are not directly comparable to the rest. Series 9 and Ultra 2 (pooled): read 8.3 ms low on average, typical miss 20.5 ms (29%), outside a ±10 ms equivalence margin; 39 people over 14 days vs a chest strap (O'Grady 2024). Series 6: 22.5 ms off in one lab night, ICC 0.67, via a third-party sleep app (Miller 2022).
Fitbit Charge 4: read 3 ms low in COPD patients (ICC 0.87) and 7 ms low in healthy adults (ICC 0.69) over up to 5 nights vs a chest strap (Hermans 2024). Versa 4: nightly median 21.7 vs 20.6 ms, rho 0.94, in an 8-person pilot, 7 with shoulder pain (Nasr 2025).
Samsung Gear Sport: 22.4 ms off asleep (r 0.78) and 32 ms awake (r 0.25) over 24 hours vs an ECG. The researchers processed raw sensor data with their own software, so this is not the HRV figure the watch shows (Sarhaddi 2022).
All brands Across 43 studies, wrist pulse-based HRV agreed acceptably with ECG at rest, but the authors warn against applying that to sleep, exercise or daily life (Xu 2026).

Studies behind this

7 sources, funding labelled

Independent

Validation of nocturnal resting heart rate and HRV in consumer wearables

Dial MB, et al. · 2025 · Physiological Reports · n=13 (536 nights)

Independent

Six wearables vs polysomnography and ECG for sleep, heart rate and HRV

Miller, et al. · 2022 · Sensors · n=53

Independent

The validity of Apple Watch Series 9 and Ultra 2 for serial measurements of heart rate variability and resting heart rate

O'Grady B, Lambe R, Baldwin M, Acheson T, Doherty C · 2024 · Sensors · n=39

Independent

Consumer wearable clinical parameters in COPD and healthy controls

Hermans F, et al. · 2024 · JMIR mHealth and uHealth · n=38

Independent

Agreement between Fitbit Versa 4 and a reference standard on time-domain nocturnal heart rate variability: a pilot study

Nasr N, Behbehani K, Wang J, et al. · 2025 · Digital Health · n=8

Independent

A comprehensive accuracy assessment of Samsung smartwatch heart rate and heart rate variability

Sarhaddi F, et al. · 2022 · PLOS ONE · n=28

Independent

Accuracy of photoplethysmography-derived pulse rate variability compared with ECG-derived HRV: systematic review and meta-analysis

Xu, et al. · 2026 · Sensors · n=43 studies

Metric 04 · Resting HR

Resting heart rate: Oura, then WHOOP.

Oura is the most accurate overnight (1.1 bpm, 1.9% error), WHOOP second (1.8 bpm) and Apple third; Garmin was 5.4 bpm off in its one lab night.

Checked Oct 2026 · Tested against chest-strap or lead II ECG

Typical error

Reads high or low

Resting heart rate overnight, how far off an ECG

Oura

1.1 bpm

WHOOP

1.8 bpm

Apple

1.5 bpm

different study

Garmin

5.4 bpm

different study

Fitbit

1.2 bpm

different study

Samsung

1.1 bpm

different study

How far resting heart rate reads high or low on average

Reads low

Reads high

Oura

-0.9 bpm

WHOOP

-1.4 bpm

Samsung

-0.4 bpm

different study

Apple

+0.5 bpm

different study

Fitbit

+4 bpm

different study

Garmin

+5.0 bpm

different study

What this means for you

Resting heart rate is the most trustworthy wearable number: Oura (1.1 bpm) and WHOOP (1.8 bpm) both stayed within 2 bpm of an ECG overnight in the largest test. If you use a Fitbit, know that its resting figure includes relaxed awake time, so it runs a few bpm high.

More data

The full picture, secondary tests and notes

The easiest number a wearable produces. Oura leads (1.1 bpm, 1.9%) and WHOOP is second (1.8 bpm, 3.0%) in the same independent overnight study; Polar sat between them on the first four hours of sleep. Apple is third (3.7 bpm lying down in the morning). Fitbit reads about 4 bpm high in healthy adults because it counts relaxed awake time, Garmin was 5.4 bpm off in one lab night and Samsung has no test of the watch feature.

Resting heart rate is where wearables are most trustworthy. Oura Ring 4 was validated on 138 nights and Gen 3 on 470. Fitbit and Samsung were not in those two studies, but older models were within about 1 bpm asleep in smaller ones (Fitbit Charge 2, Samsung Gear Sport). Oura Ring 5 has no validation of any kind.

11.1 bpm

off while awake versus 1.1 bpm asleep: the same Samsung watch, the same 24-hour recording (Sarhaddi 2022). Night is easy, day is hard.

Oura Ring 4: 1.1 bpm, 1.9% (138 nights); Gen 3: 1.0 bpm, 1.7% (470 nights), Dial 2025. Gen 2: 1.8 bpm in one night (Miller 2022).
WHOOP 4.0: 1.8 bpm, 3.0% (288 nights, Dial 2025); 1.3 bpm asleep and 1.7 bpm awake over 5 days (Zia Ur Rehman 2024). 3.0: 0.7 bpm in a lab that receives WHOOP funding (Miller 2022).
Fitbit Charge 2 overnight: 0.09 bpm against a 3-lead ECG (Haghayegh 2019) and 0.9 bpm in a home sleep study (Stucky 2021).
Samsung Gear Sport: 1.1 bpm asleep but 11.1 bpm awake in the same recording (Sarhaddi 2022).
Apple Series 6: 1.5 bpm in one night (Miller 2022).
Garmin Forerunner 245: 5.4 bpm in one night (Miller 2022). The Fenix 6 was left out of Dial 2025 because Garmin does not say which 30 minutes it uses.

Studies behind this

8 sources, funding labelled

Independent

Validation of nocturnal resting heart rate and HRV in consumer wearables

Dial MB, et al. · 2025 · Physiological Reports · n=13 (536 nights)

Independent

Six wearables vs polysomnography and ECG for sleep, heart rate and HRV

Miller, et al. · 2022 · Sensors · n=53

Independent

A comprehensive accuracy assessment of Samsung smartwatch heart rate and heart rate variability

Sarhaddi F, et al. · 2022 · PLOS ONE · n=28

Independent

Assessment of physiological signals from photoplethysmography sensors compared to an electrocardiogram sensor in daily life

Zia Ur Rehman R, et al. · 2024 · Sensors · n=25

Independent

Accuracy of Fitbit Charge 2 heart rate during sleep

Haghayegh S, et al. · 2019 · Chronobiology International · n=35

Independent

Validation of Fitbit Charge 2 sleep and heart rate in shift workers

Stucky B, et al. · 2021 · Journal of Medical Internet Research · n=59

Independent

Consumer wearable clinical parameters in COPD and healthy controls

Hermans F, et al. · 2024 · JMIR mHealth and uHealth · n=38

Independent

Assessment of Fitbit Charge 4 for sleep stage and heart rate monitoring against polysomnography and during home monitoring in Huntington's disease

Doheny EP, Renerts K, Braun A, et al. · 2024 · Journal of Clinical Sleep Medicine · n=10

Metric 05 · Active HR

Heart rate during activity: Apple edges Garmin and Fitbit.

Checked Oct 2026 · Tested against chest strap or 12-lead ECG

Daytime

By sport

Heart rate during daytime activity

Fitbit

5.5%

Garmin

6.3%

Apple

7.3%

Oura

15.0%

WHOOP

8.5%

different study

Not in this test:

Samsung

Same watch, same people, real training

Running

1.2%

Cycling

8.1%

Tennis

8.9%

Orienteering

9.5%

Badminton

16.2%

Soccer

17.5%

What this means for you

Apple, Garmin and Fitbit are all within a few bpm on steady cardio; the brand matters less than the sport. Steady running reads well, racquet and field sports read worst. Two free fixes: wear the watch three finger-widths above the wrist bone and tighten the strap.

Most Accurate Heart Rate Wearable

Dedicated tool · 10 devices · tiers, not rankings

Open

More data

The full picture, secondary tests and notes

Apple is first: lowest error of four watches cycling (2.5 bpm), bias zero in a second cycling test and near-zero pooled bias across 22 studies. Garmin is second (2.9 bpm cycling, 6.3% in the ten-device test) and Fitbit third (best of ten with the Charge 6 at 5.5%, but last of four cycling with the Sense 2). Samsung is fourth; its sub-2% result was heart patients on a stress test. The order is not firm: the top four are within 2 to 7%. Arm movement, not intensity, is what breaks the reading.

The failure mode is irregular arm movement, not intensity: steady running is easy, racquet and field sports are worst. The WHOOP 99.7% and Oura 99% accuracy claims come from sleep data. Two free fixes: wear the watch three finger-widths above the wrist bone and tighten the strap.

-0.27 bpm

Apple Watch, 22 studies. Pooled bias near zero; Series 6 onward stay within -3.7 to +2.6 bpm (Lambe 2026).

-3.39 bpm

Fitbit, 32 studies. Reads low on average, with wide limits of -24.3 to +17.5 bpm (Chevance 2022).

Fitbit Charge 6: 5.5%, Tier 1. The Inspire 3 in the same test: 16.5%, Tier 4 (Gielen 2026). Model matters more than brand.
Garmin Vivoactive 5: 6.3%, Tier 1. Vivosmart 5: 8.1%, Tier 2 (Gielen 2026).
Apple Watch SE: 7.3%, Tier 2 (Gielen 2026). Apple Watch 7: ICC 0.998 in a hospital stress test (Kim 2023).
WHOOP 4.0: 8.5% on a treadmill against a 12-lead ECG, with 13% of readings missing (Van Oost 2025).
Samsung Galaxy Watch 4: under 2% in heart patients on a hospital treadmill, falling to ICC 0.867 above 160 bpm (Kim 2023).

Studies behind this

6 sources, funding labelled

Independent

Ten-device wrist heart rate validation

Gielen J, et al. · 2026 · JMIR Formative Research · n=45 (10 sessions per device)

Independent

Five wearables vs 12-lead ECG during exercise

Van Oost L, et al. · 2025 · Sensors · n=24

Independent

Apple Watch 7 and Galaxy Watch 4 vs 12-lead ECG during cardiopulmonary exercise testing

Kim J, et al. · 2023 · Annals of Rehabilitation Medicine · n=44 cardiac patients

Independent

Fitbit Charge 4 heart rate across real training sessions and sports

Ceugniez M, et al. · 2025 · JMIR mHealth and uHealth · n=26

Independent

The accuracy of Apple Watch measurements: a living systematic review and meta-analysis

Lambe R, et al. · 2026 · npj Digital Medicine · n=82 studies

Independent

Accuracy and precision of energy expenditure, heart rate and steps measured by Fitbit

Chevance G, et al. · 2022 · JMIR mHealth and uHealth · n=32 heart rate studies

Metric 06 · SpO2

SpO2: least reliable when it counts.

Apple is the most accurate (2.2 points off in a low-oxygen test), Samsung second (3.8, reads low) and Garmin a distant third. Every watch is least reliable when oxygen is truly low.

Checked Oct 2026 · Tested against arterial blood or hospital oximeter

Typical error

Reads high or low

How far off a medical reference

Apple

2.2%

Samsung

3.8%

Garmin

5.8%

different study

Not in this test:

Oura

WHOOP

Fitbit

How far blood oxygen reads high or low on average

Reads low

Reads high

Apple

-0.1 pts

Samsung

-2.6 pts

Fitbit

+2 pts

different study

Garmin

-5.5 pts

different study

Not in this test:

Oura

WHOOP

What this means for you

A normal reading is not reassurance if you feel unwell, because no watch has proven reliable when oxygen is truly low. No SpO2 feature on these brands is FDA-cleared.

More data

The full picture, secondary tests and notes

Apple leads: Arms 2.2 in a low-oxygen lab test, 2.4 against arterial blood in 112 patients, pooled bias near zero across nine studies. Samsung is second (Arms 3.8, reads 2.6 low) and Garmin third by default (3.5 against arterial blood, fails the ISO limit against a finger oximeter). Fitbit has two small, poor tests; Oura and WHOOP have none. When oxygen is truly low no watch has proven reliable, and 10 to 45% of readings simply fail.

At normal levels the best watches land within about 2 points. When oxygen is truly low, the range that matters, an Apple Watch over-read most readings in one test, so a normal number is not reassurance. No SpO2 feature on any of these brands is FDA-cleared.

85%

of Apple Watch Series 7 readings were too high when true oxygen was below 88%, in a test against arterial blood (Jiang 2026, 9 people).

Up to 92%

of readings failed with moderate movement at simulated altitude, across devices including Apple and Garmin (Hearn 2023).

Apple Series 8 in the lab test: 2.2% at all levels, 2.5% at 80% oxygen or below (Walzel 2023). Series 7: 2.2% in clinic patients at rest with 11% of readings failing (Jiang 2023), and 4.6% against arterial blood (Jiang 2026).
Samsung Galaxy Watch 5 in the lab test: 3.8% at all levels, 2.7% at 80% or below, reading low (Walzel 2023). Galaxy Watch 4: 1.8% against arterial blood in a Samsung-funded study (Browne 2025).
Garmin Venu 2s: 5.8% at rest with 14% of readings failing; Fenix 6 Pro: 28% of readings failed (Jiang 2023). Venu 3 / Venu 4: 2.96% overnight in 29 ward patients, a preprint (Knevels 2026).
Fitbit Charge 4 read about 2 points high overnight in COPD patients, ICC 0.32 (Hermans 2024).
Oura Oura reports "30% more accurate" than Gen 3 from its own 60-person test; no absolute error published.
WHOOP No published SpO2 accuracy study found.
Apple US watches have computed blood oxygen on the paired iPhone since August 2025. That pipeline has no independent validation.

Studies behind this

12 sources, funding labelled

Independent

Evaluation of leading smartwatches for the detection of hypoxemia

Walzel S, et al. · 2023 · Sensors · n=18

Independent

Accuracy of blood oxygen saturation in common consumer smartwatches

Jiang Y, et al. · 2023 · PLOS Digital Health · n=49

Independent

Wearable pulse oximetry during controlled hypoxia induction

Jiang Y, et al. · 2026 · JMIR Formative Research · n=9 (308 blood samples)

Samsung-funded

Smartphone and smartwatch pulse oximetry during hypoxia laboratory testing

Browne SH, et al. · 2025 · Sensors · n=12 in the watch arm

Independent

Smartwatch vs polysomnography for overnight oximetry and sleep apnea evaluation

Browne SH, et al. · 2024 · Journal of Clinical Sleep Medicine · n=51

Independent

Garmin Fenix 5X Plus oxygen saturation at 4,559 m

Schiefer LM, et al. · 2021 · Sensors · n=13

Independent

Garmin Venu heart rate, respiratory rate and SpO2 in a respiratory ward

Knevels R, et al. · 2026 · Research Square (preprint) · n=29

Independent

Consumer wearable clinical parameters in COPD and healthy controls

Hermans F, et al. · 2024 · JMIR mHealth and uHealth · n=38

Independent

Fitbit Sense 2 in hospitalized general medicine patients

Lee, et al. · 2026 · International Journal of Medical Informatics · n=33

Independent

Arterial oxygen saturation from wearables under varying conditions

Hearn, et al. · 2023 · Aerospace Medicine and Human Performance · n=10

Independent

The accuracy of Apple Watch measurements: a living systematic review and meta-analysis

Lambe R, et al. · 2026 · npj Digital Medicine · n=82 studies

Apple-funded

Blood Oxygen app on Apple Watch (white paper)

Apple · 2022 · Apple · n=50

Metric 07 · Steps

Step counting: Garmin leads in daily life.

Garmin counts steps most accurately: 3% over in daily life across 59 studies and 1 to 2% in the lab. Apple is second (7%), Samsung third. Fitbit and especially Oura overcount in daily life.

Checked Oct 2026 · Tested against hand count (lab) or research pedometer (daily life)

Daily life

Lab walking

Steps over whole days of normal life

Average over or under

Typical miss

Reads low

Reads high

Apple

+2% · ±6%

Garmin

±18%

Fitbit

+18%

Oura

+19% · ±50%

Samsung

±10.5%

Not in this test:

WHOOP

Against a hand or video count

Typical miss

Reads low

Reads high

Fitbit

±3.6%

Garmin

±4.6%

Apple

±9.3% slow walk

Not in this test:

Oura

WHOOP

Samsung

What this means for you

Garmin and Apple are the counts to trust day to day (3% and 7% over, pooled). Fitbit runs about 15% high in daily life. Oura overcounted by 2,000 to 3,800 steps a day before its 2025 update, so use its count as a trend.

Step Count Accuracy

Dedicated tool · Lab vs daily life

Open

More data

The full picture, secondary tests and notes

Garmin is first: 3% over in daily life pooled across 59 studies, and 1 to 2% on a treadmill in 258 adults. Apple is second (7% pooled, 6.4% in a 36-adult study, within 3.4% at every treadmill speed) and Samsung third (10.5% in daily life). Fitbit is 3.6% on a normal lab walk but overcounts by 15% in daily life, which is how people use it. Oura was 50 to 69% off before its May 2025 algorithm change, with nothing published since.

One podium: Garmin first, Apple second, Samsung third, set by the pooled free-living result across 59 studies (Garmin +3%, Apple +7%, Fitbit +15%) and the lab tests. On a normal treadmill walk Fitbit and Garmin land under 5%; Apple's recent lab tests used slow or assisted walking, where every wrist device struggles. Apple beat Samsung in the same 24-hour study (6.4% vs 10.5%). Both Gen 2 studies of Oura found it overcounting by 1,400 to 2,100 steps a day; its 2025 "Real Steps" algorithm has no published validation. No Fitbit newer than the Sense or Inspire 2, no Pixel Watch and no WHOOP has peer-reviewed step data.

Apple Daily life: Series 6 +2.1% over 3 weeks (Miwa 2026) and 6.4% typical miss over 24 hours (Hong 2024). Lab: Series 5 ±9.3% on slow and shuffle walks (Rowe 2025); Series 8 -16% at a very slow walk and -36% with a rolling walker in adults 51 to 80 (Özel 2026).
Garmin Daily life: Vivofit (2014 model) 17.8% typical miss at home vs a pedometer (An 2017, per the Evenson & Spade 2020 review of 32 Garmin studies). Short walking and jogging bouts: Vivosmart HR+ within 10% over ground and 5% on a treadmill (Montes 2020). Lab: Vivoactive 4 ±4.6% (de Leon 2026); Vivosmart HR 0.6 to 1.3% at 3.2 to 4.8 km/h, undercounting at slow speeds (Svarre 2020); Fenix 6 -15% on a treadmill-and-chores protocol (Rider 2025).
Fitbit Daily life: Sense +18% (Miwa 2026); Charge 2 and Alta 17 to 36% typical miss (Giurgiu 2023). Lab: Inspire 2 ±3.6% (Cheung 2025).
Oura Gen 2: +2,124 steps a day, 50% typical daily miss (Kristiansson 2023, corrected); +1,416 a day against an ActiGraph (Niela-Vilén 2022). Gen 3: -6.2% in only 5 people (Miwa 2026). The 2025 Real Steps update has no published validation. No lab step data exists.
Samsung Galaxy Watch 4: 10.5% typical miss over 24 hours of daily life, r 0.82, against a thigh-worn ActivPAL; Apple Watch 6 in the same study: 6.4%, r 0.99 (Hong 2024). No lab study.
WHOOP No published step study.

Studies behind this

13 sources, funding labelled

Independent

Apple Watch 6 vs Galaxy Watch 4: a validity study of step-count estimation in daily activities

Hong KR, Hwang IW, Kim HJ, Yang SH, Lee JM · 2024 · Sensors · n=104

Independent

Consumer wearables vs a research-grade monitor for physical activity in free living

Miwa T, et al. · 2026 · PLOS ONE · n=36

Independent

Apple Watch step counting during slow and shuffling gait

Rowe G, et al. · 2025 · PLOS ONE

Independent

Garmin Vivoactive 4 heart rate, calories and steps on a treadmill

de Leon MF, et al. · 2026 · Applied Sciences · n=18

Independent

Four smartwatches in laboratory and field conditions

Rider BC, et al. · 2025 · Journal for the Measurement of Physical Behaviour · n=12

Independent

Fitbit Inspire 2 treadmill step accuracy

Cheung YS, et al. · 2025 · Physiotherapy Practice and Research · n=30

Independent

Why your smartwatch may be misleading your doctor: mobility aids and wearable accuracy in older adults

Özel A, et al. · 2026 · PeerJ · n=42 adults aged 51 to 80

Independent

Validity of wearable monitors and smartphone applications for measuring steps in semi-structured and free-living settings

Giurgiu M, et al. · 2023 · Technologies

Independent

Validation of Oura ring energy expenditure and steps in laboratory and free-living (correction 10.1186/s12874-023-02029-w)

Kristiansson E, Fridolfsson J, Arvidsson D, Holmäng A, Börjesson M, Andersson-Hall U · 2023 · BMC Medical Research Methodology

Independent

The validity of activity trackers is affected by walking speed: Garmin VivosmartHR and StepWatch3

Svarre FR, Jensen MM, Nielsen J, Villumsen M · 2020 · PeerJ · n=30

Independent

Comparison of Oura smart ring against ActiGraph accelerometer for physical activity and sedentary time in a free-living context

Niela-Vilén H, Azimi I, Suorsa K, et al. · 2022 · Computers, Informatics, Nursing · n=42

Independent

How valid are wearable physical activity trackers for measuring steps?

An HS, Jones GC, Kang SK, Welk GJ, Lee JM · 2017 · European Journal of Sport Science · n=35

Independent

Step count reliability and validity of five wearable technology devices while walking and jogging in a free motion setting and on a treadmill

Montes J, Tandy R, Young J, Lee SP, Navalta JW · 2020 · International Journal of Exercise Science · n=40

Metric 08 · Calories

Calories burned: Fitbit is least bad; none is reliable.

The least accurate number on any wearable. Fitbit is least bad, Samsung second, Apple third, and every brand is more than 10% off in every head-to-head test.

Checked Oct 2026 · Tested against indirect calorimetry (lab gas analysis)

Workouts

Reads high or low

Daily totals

Workout calories, typical miss either way (the studies give no direction)

Cycling

Weight training

Reads low

Reads high

Fitbit

±32%

±42%

Apple

±36%

±112%

Samsung

±67%

±85%

Garmin

±78%

±115%

Not in this test:

Oura

WHOOP

Average over or under the real burn, same 62 people

Endurance

Weight training

Reads low

Reads high

Fitbit

+21%

+4%

Apple

+14%

+112%

Samsung

+9%

+83%

Garmin

+30%

+116%

Not in this test:

Oura

WHOOP

Calories across whole days of normal life

Whole-day burn

Activity calories only

Reads low

Reads high

Fitbit

±9 to 13%

+139%

Oura

±13%

Not tested

Apple

Not tested

+26%

Not in this test:

Garmin

WHOOP

Samsung

What this means for you

Fitbit is the least bad, Samsung is the most accurate on steady running and walking, and nobody is within 10%. Weight-training errors run from 42% (Fitbit) to 115% (Garmin). Error grows with body fat. Treat the number as a trend, not a target.

Calorie Burn Accuracy Calculator

Dedicated tool · Calculator · by device and activity

Open

More data

The full picture, secondary tests and notes

The least accurate number on any wearable. Across four head-to-head lab tests on the same people, Fitbit wins three of five comparisons and has the smallest worst case (42% on weights, where it reads only 4% high); Samsung is the most accurate on steady running and walking but roughly doubles true burn on weights; Apple and Garmin are off by 112% and 115% on weights. Every brand is more than 10% off in every test. Oura has never been tested against lab calorimetry and WHOOP has no calorie study.

Calories are the least accurate number on any wearable, so the podium separates bad from worse: Fitbit wins three of five head-to-head comparisons (max treadmill 14%, cycling 32%, weights 42%, reads only 4% high on weights), Samsung is second (9% on a max treadmill and 0.5% walking, but 67% cycling and 85% weights), Apple third (26%, 36%, 112%). Garmin is 78% off cycling and 115% on weights. Error also grows with body fat. Treat the number as a trend, not a target.

Fitbit In cycling, 7 implausible readings (about 13% of Fitbit trials) were removed; with them, Fitbit had the largest error of all four. Daily calories: the Alta HR was 9 to 13% off doubly labelled water, the gold standard (Shook 2022).
Garmin Vivoactive 4 on a treadmill run: 19.1% (de Leon 2026). Forerunner 935: 15.3% (Uphill 2024). Garmin's direction is not consistent across studies.
Samsung Galaxy Watch 6 and 7: 10 to 13% in a Samsung-funded study (Ferreira 2026).
Apple Apple Watch 5: 11.3% (Uphill 2024). A 2026 living review could not pool Apple calorie results: errors ranged from 9.7% to 152% (Lambe 2026).
Apple Apple's own white paper reports -2% to +5% across 11 workout types. Independent studies have not reproduced this.
Oura Gen 2 in the lab: 21.1% (Kristiansson 2023). Oura changed its active-calorie algorithm in May 2025; nobody has tested the new one.
WHOOP No published calorie validation of any kind.

Max treadmill test, a conference abstract with about 20 different people per watch, so not a ranking (Perez 2025): Galaxy Watch7 9.4%, Fitbit Charge 6 14%, Garmin Forerunner 265S 15%, Apple Ultra 2 26%.

Studies behind this

10 sources, funding labelled

Independent

Four smartwatches vs indirect calorimetry during cycling

Kostrna, et al. · 2026 · PLOS ONE · n=58

Independent

Four smartwatches during resistance exercise

Yun, Ko, Lee · 2026 · Bioengineering · n=70

Independent

Four smartwatches during endurance and resistance exercise vs indirect calorimetry

Lee TH, et al. · 2026 · Sensors · n=62

Independent

Five wrist wearables with sex-specific energy expenditure algorithms

Perez, Davis, Wong, Navalta · 2025 · Int J Exercise Science Conference Proceedings (abstract) · n=97 (about 20 per device)

Independent

Fitbit Alta HR vs doubly labelled water

Shook RP, et al. · 2022 · Journal of Nutrition · n=24

Independent

Oura Gen 2 and Polar Vantage vs research monitors in free living

Henriksen A, et al. · 2022 · JMIR Formative Research

Independent

Consumer wearables vs a research-grade monitor for physical activity in free living

Miwa T, et al. · 2026 · PLOS ONE · n=36

Samsung-funded

Galaxy Watch 6 and 7 during intermittent running

Ferreira, et al. · 2026 · JMIR Formative Research · n=148

Apple-funded

Heart Rate, Calorimetry, and Activity on Apple Watch (white paper)

Apple · 2024 · Apple

Independent

Accuracy in wrist-worn devices for heart rate and energy expenditure

Shcherbina A, et al. · 2017 · Journal of Personalized Medicine · n=60

Metric 09 · VO2 max

VO2 max: Garmin leads, with a chest strap.

Garmin has the most validation at about 7% error. Apple reads low in all three independent studies, and Fitbit's Pixel Watch 3 came in at 8.7% in a Google-funded test.

Checked Oct 2026 · Tested against lab metabolic cart (CPET)

Typical error

How far off a lab VO2 max test

Garmin

6.7%

Fitbit

8.7%

Apple

13.2%

Not in this test:

Oura

WHOOP

Samsung

What this means for you

Garmin with a chest strap is the estimate to trust (about 7% off); wrist-only Garmin in older adults was 56% off. Estimates from a real outdoor run beat ones from resting heart rate. Apple reads low, most in fit women. Use it as a trend.

VO2 Max Accuracy

Dedicated tool · 9 devices · by method

Open

More data

The full picture, secondary tests and notes

Garmin has the most independent validation at about 7% error. Apple has three independent studies and underestimates in all of them. Fitbit's Pixel Watch 3 came in at 8.7% in a Google-funded test; its older Charge 2 studies tested a method Google has retired. Oura, WHOOP and Samsung have vendor claims only.

Method beats brand: estimates from a real outdoor run show about zero group bias, estimates from resting heart rate overestimate by about 2 mL/kg/min (INTERLIVE 2022). Accuracy degrades in highly trained users, and the one elite study that avoided this used a chest strap. Use it as a trend, not a number.

Apple Reads low in every study: Series 10 −6.3 (13.2% typical error, Lambe 2026), Series 9 and Ultra 2 −6.1 (13.3%, Lambe 2025), Series 7 −4.5 (15.8%, Caserman 2024).
Garmin Forerunner 245: 6.7% overall in 35 endurance runners, 2.8 to 4.1% in moderately trained but 9.4 to 10.4% in highly trained (Engel 2026). Fenix 6: 7.1% (Carrier 2025); 6.9% with a chest strap in elite athletes (Carrier 2023). No newer Garmin has been studied.
Fitbit Pixel Watch 3: 8.7%, reading 0.5 low (Lambe 2026, Google-funded, 26 adults), on the outdoor-run method Google now ships. Charge 2: 9.1%, reading 0.3 high (Klepin 2019, 60 people); +2.6 high (Freeberg 2019). Both Charge 2 studies tested an algorithm Google retired in May 2026.
WHOOP Company figures only: about 8% typical error, off by 3.3 to 3.7 mL/kg/min on average (r 0.90) in WHOOP's own test of 248 people against a lab gas analyser. No independent check. VO2 max is on WHOOP 5.0 and MG only.
Oura No independent study and no published accuracy figure. Oura states its 6-minute walk test estimate is less accurate than a lab test and weaker for athletes and at altitude.
Samsung Company-funded testing only: r 0.82 against lab equipment (Samsung-funded University of Michigan testing). The only independent Galaxy Watch study checked heart rate during a max test, not VO2 max.

Fitbit now derives VO2 max from outdoor GPS runs, a method with no published validation.

Studies behind this

10 sources, funding labelled

Independent

Forerunner 245 VO2 max in highly vs moderately trained runners

Engel FA, et al. · 2026 · European Journal of Applied Physiology · n=35

Independent

Validation of aerobic capacity (VO2max) and pulse oximetry in wearable technology

Carrier B, et al. · 2025 · Sensors · n=19

Independent

Garmin Fenix 6 with chest strap in an athletic sample

Carrier B, et al. · 2023 · Technologies · n=20

Independent

Apple Watch VO2 max estimation

Caserman P, et al. · 2024 · JMIR Biomedical Engineering

Independent

Validation of Apple Watch VO2 max estimates

Lambe RF, et al. · 2025 · PLOS ONE

Independent

Accuracy of VO2 max estimates from Apple Watch Series 10

Lambe R, et al. · 2026 · Mayo Clinic Proceedings: Digital Health · n=40

Google-funded

Validity of the Google Pixel Watch 3 for estimating VO2 max and heart rate against a maximal treadmill test

Lambe R, et al. · 2026 · PLOS ONE · n=26

Independent

Validity of cardiorespiratory fitness measured with Fitbit compared to VO2max

Klepin K, et al. · 2019 · Medicine & Science in Sports & Exercise · n=60

Independent

Fitbit Charge 2 VO2 max estimation

Freeberg KA, et al. · 2019 · mHealth

Independent

Validity of consumer wearable VO2 max estimates: meta-analysis

Molina-Garcia P, et al. (INTERLIVE) · 2022 · Sports Medicine · n=Meta-analysis

Metric 10 · Skin temp

Skin temperature: no independent sensor test.

No independent study has tested any brand's temperature sensor.

Checked Oct 2026 · Tested against research-grade iButton sensors

Wearable	Sensor test	Cycle tracking study	Grade
Apple	89% within 2 days Apple-funded, prototype
Oura	r² above 0.92 worn Own 2020 test	96.4% detected Oura staff	Company claim only
Garmin
WHOOP
Fitbit
Samsung	78% within 3 days Samsung-funded

No independent study has tested any brand's temperature sensor. Cycle figures use different definitions and are not comparable across brands.

What this means for you

Use skin temperature for nightly change against your own baseline, not as a thermometer. For cycle tracking, the FDA-cleared part is the Natural Cycles app, not the ring or watch.

More data

The full picture, secondary tests and notes

No independent study has tested any brand's temperature sensor. Oura is the only brand that has published a sensor test, and it ran that test itself.

There is no winner because there is no independent comparison. Oura's lab figures (r² above 0.99, within 0.36°C) come from rings tested on their own in water baths. Apple reports nightly change from baseline and states it is not a thermometer. For cycle tracking, the FDA-cleared part is the Natural Cycles app, not the ring or watch.

Studies behind this

4 sources, funding labelled

Oura-funded

How accurate is Oura's temperature data? (internal study, not peer reviewed)

Oura · 2020 · Oura blog · n=16 people + lab bath test

Oura-funded

Oura Ring as a tool for ovulation detection

Thigpen N, et al. · 2025 · Journal of Medical Internet Research · n=964 users, 1,155 cycles

Apple-funded

Wrist temperature algorithms for ovulation and next-period prediction

Wang Y, et al. · 2025 · Human Reproduction · n=260 people, 889 cycles

Samsung-funded

Galaxy Watch 5 wrist temperature for ovulation prediction

Kim SE, et al. · 2026 · Biosensors and Bioelectronics · n=85 women, 225 cycles

Metric 11 · Breathing

Breathing rate: only two brands qualify.

Fitbit is the only brand with an independent test of the shipped feature (0.3 to 1 breath a minute high); WHOOP is second on a peer-reviewed company-funded test. No third qualifies.

Checked Oct 2026 · Tested against nasal airflow, chest bands or impedance

Breathing rate

Overnight breathing rate, breaths per minute off

Independent test

Company-run test

Fitbit

0.3 to 1/min

0.5/min

Garmin

1.5/min low

Not tested

Oura

Not tested

0.7/min

Samsung

Not tested

1.1/min

Not in this test:

Apple

WHOOP

What this means for you

Overnight averages on a Fitbit or WHOOP are the best-supported use. Garmin's breathing rate failed its hospital test and Apple only measures during sleep. Treat single readings with caution.

More data

The full picture, secondary tests and notes

Only two brands qualify. Fitbit is first: an independent overnight test put the Charge 4 0.3 to 1 breath a minute high, and Fitbit's own test agrees (RMSE 0.65). WHOOP is second on a WHOOP-funded but peer-reviewed sleep-lab test of the Strap 2.0 (+0.1 a minute). Garmin's shipped feature failed its hospital test (r 0.28), Apple gave no readings in two hospital studies, Oura's 0.71 is a blog post and Samsung's 1.13 is a research algorithm.

Fitbit first, WHOOP second, no third. Company-run tests in sleeping, mostly healthy adults report errors under 1 to 2 breaths a minute, and Fitbit's independent COPD-and-healthy test agrees (0.3 to 1 high). A Garmin barely followed real breathing in hospital patients and an Apple Watch gave no readings. Samsung's often-quoted figure came from a research algorithm the watch did not offer to users. Overnight averages are the best-supported use.

Fitbit Charge 4: reads 0.3 high in COPD and 1 high in healthy adults, ICC 0.77 to 0.84, against an ECG-derived reference (Hermans 2024). Company test: 0.46 average error (Natarajan 2021).
Garmin Fenix 6 Pro: correlation 0.28 with the hospital monitor, reading 1.5 low (Helmer 2024). Venu 3 / Venu 4: 4.2 off, 47% of readings within 3, acceptable only up to 22 breaths a minute (Knevels 2026, preprint).
Samsung Research algorithm: 1.13 RMSE overall, 0.46 in milder sleep apnea but 2.17 in severe apnea (Jung 2023, Samsung-funded).
Apple No readings in two hospital studies (Helmer 2024, AbiMansour 2024). A 2026 living review lists respiratory rate as yet to be validated.
WHOOP Strap 2.0: 1.8% bias only (Berryhill 2020, WHOOP-funded). Nothing for WHOOP 4.0 or 5.0.

Studies behind this

8 sources, funding labelled

Samsung-funded

Validating a consumer smartwatch for nocturnal respiratory rate

Jung H, et al. · 2023 · Sensors · n=195 sleep-clinic patients

Fitbit-funded

Measurement of respiratory rate using wearable devices

Natarajan A, et al. · 2021 · npj Digital Medicine · n=28 people, 52 nights

WHOOP-funded

Effect of wearables on sleep in healthy individuals: validation study

Berryhill S, et al. · 2020 · Journal of Clinical Sleep Medicine · n=32

Oura-funded

How accurate is Oura's respiratory rate? (not peer reviewed)

Oura · 2020 · Oura blog · n=43

Independent

Continuous vital sign monitoring in post-operative patients with consumer fitness trackers

Helmer P, et al. · 2024 · Digital Health · n=12 in the Garmin arm

Independent

Garmin Venu heart rate, respiratory rate and SpO2 in a respiratory ward

Knevels R, et al. · 2026 · Research Square (preprint) · n=29

Independent

Consumer wearable clinical parameters in COPD and healthy controls

Hermans F, et al. · 2024 · JMIR mHealth and uHealth · n=38

Independent

Consumer wearables for continuous monitoring during endoscopy

AbiMansour JP, et al. · 2024 · Frontiers in Digital Health · n=292 procedures

Metric 12 · FDA-cleared

FDA-cleared features: cleared is not caught.

Apple has the most cleared features (5) and Oura has none, but cleared does not mean every case is caught.

Checked Oct 2026 · Tested against FDA 510(k) and De Novo decisions

Feature	Apple	Oura	Garmin	WHOOP	Fitbit	Samsung
ECG / AFib check	Series 4+	Select watches	MG only	Sense, Charge 5/6	Watch4+
Irregular rhythm alerts	Most models
AFib History
Sleep apnea alerts	S9+, Ultra 2+, SE 3	Watch and Ring
Hypertension alerts	S9+, Ultra 2+
Blood pressure (wellness, not cleared)	Wellness Ring 5, no numbers	Wellness MG	Wellness Pixel	Wellness
Cleared features	5	0	1	1	2	3

Apple

cleared

ECG / AFib check Series 4+
Irregular rhythm alerts
AFib History
Sleep apnea alerts S9+, Ultra 2+, SE 3
Hypertension alerts S9+, Ultra 2+

Oura

cleared

Blood pressure (wellness, not cleared) Ring 5, no numbers

Garmin

cleared

ECG / AFib check Select watches

WHOOP

cleared

ECG / AFib check MG only
Blood pressure (wellness, not cleared) MG

Fitbit

cleared

ECG / AFib check Sense, Charge 5/6
Irregular rhythm alerts Most models
Blood pressure (wellness, not cleared) Pixel

Samsung

cleared

ECG / AFib check Watch4+
Irregular rhythm alerts
Sleep apnea alerts Watch and Ring
Blood pressure (wellness, not cleared)

What this means for you

An alert is worth acting on, but no alert is not an all-clear. Only Apple's hypertension alert is cleared for blood pressure, and it gives no numbers.

More data

The full picture, secondary tests and notes

Apple has the most cleared features (5), then Samsung (3), Fitbit (2), Garmin and WHOOP MG (1 each, ECG). Oura has none. Clearance means a claim was reviewed, not that the alert catches everything.

ECG apps from all five cleared brands score 96% to 100% on recordings they could classify, because inconclusive ones are excluded (2.9% for Samsung up to 11.5% for Garmin), so do not rank brands on ECG. Nothing here is "FDA-approved"; these are clearances for specific features. Only Apple's hypertension alert is cleared for blood pressure, and it gives no numbers.

Cleared does not mean it catches everything

Share of real cases each alert flagged in the maker's own FDA study. Different features and studies, not a brand ranking.

Apple hypertension

92.3% specificity

41.2%

Apple sleep apnea

98.5% specificity

66.3%

Samsung irregular rhythm

98.8% specificity

68%

Pixel Watch loss of pulse

99.97% specificity

69.3%

Samsung sleep apnea v2.0, ring

95.9% specificity

77.9%

Samsung sleep apnea v2.0, watch

90.3% specificity

90.8%

Studies behind this

14 sources, funding labelled

FDA record

Apple Hypertension Notification Feature

K250507 · 2025

FDA record

Apple Sleep Apnea Notification Feature

K240929 · 2024

FDA record

Apple Irregular Rhythm Notification Feature

DEN180042 · 2018

FDA record

Apple ECG App

DEN180044 · 2018

FDA record

Apple AFib History Feature

K213971 · 2022

FDA record

Samsung ECG Monitor App

K201168 · 2020

FDA record

Samsung Irregular Heart Rhythm Notification

K230292 · 2023

FDA record

Samsung Sleep Apnea Feature (first De Novo)

DEN230041 · 2024

FDA record

Samsung Sleep Apnea Feature v2.0 (Galaxy Watch and Galaxy Ring)

K261011 · 2026

FDA record

Fitbit ECG App

K200948 · 2020

FDA record

Fitbit Irregular Rhythm Notifications

K212372 · 2022

FDA record

Pixel Watch Loss of Pulse Detection

K242967 · 2025

FDA record

Garmin ECG App

K221774 · 2023

FDA record

WHOOP MG Heart Screener (ECG)

K243236 · 2025

Keep exploring

Related tools.

More free, evidence-based tools to get the most out of your wearable.

What moves accuracy

Wearable Accuracy: What Actually Affects It

Tattoos, hairy arms, cold hands, strap tightness, which wrist. What the studies found and the free fixes that work tonight.

Wearables · 26 answers

Open

Sensor & software focus

Hardware & Software Differences

What actually makes Garmin, Whoop, Oura, Apple Watch and Fitbit different, sensor by sensor.

Wearables · 6 devices

Open

Error vs chest strap

Fitbit

Apple

Oura

Xiaomi

Most Accurate Heart Rate Wearable

10 wearables ranked against an ECG or chest strap, and why arm motion is what breaks them.

Wearables · 10 devices

Open

FAQ

Common questions.

Which wearable is the most accurate overall?

No single wearable is most accurate across all metrics. Oura Ring is the most accurate for everything measured overnight: sleep stages (kappa 0.65, Ring 3), sleep timing (total sleep within 2 minutes of a lab study), overnight HRV (6% error, agreement 0.99, Ring 4) and resting heart rate (1.1 bpm). Apple Watch is the most accurate for heart rate during exercise (lowest error of four watches cycling, 2.5 bpm) and blood oxygen (2.2 points off in a low-oxygen test), and has the most FDA-cleared features (five). Garmin is the most accurate for steps (3% over in daily life, pooled across 59 studies) and VO2 max (6.7% error with a chest strap). Fitbit is the least bad for calories and the only brand with an independent breathing rate test. WHOOP is second for HRV and resting heart rate. Samsung is second for blood oxygen and calories. These rankings are for the models tested, mostly one to three generations old; the newest models have little or no published validation.

How accurate is Oura Ring for sleep tracking?

It depends on who is wearing it. In an Oura-funded study of screened healthy sleepers (Robbins 2024, n=35) the Gen 3 ring reached a kappa of 0.65 against a lab sleep study. In independent clinical samples it falls to 0.35 (Lee 2023) and 53% stage accuracy in sleep-clinic patients (Herberger 2025). Oura is strongest at total sleep time, with a bias of about -3 minutes. Oura Ring 5 has no published validation.

Is Apple Watch or Garmin more accurate for heart rate?

Apple, by a small margin. In the one cycling test that put both on the same people (Kostrna 2026) the Apple Watch Series 8 had the lowest error of four watches (2.5 bpm) and the Garmin Forerunner 955 was second (2.9 bpm); across 22 studies Apple's pooled bias is close to zero (-0.27 bpm, Lambe 2026). In a ten-device test against a chest strap (Gielen 2026) the Garmin Vivoactive 5 came in at 6.3% and the budget Apple Watch SE at 7.3%, with the Fitbit Charge 6 best at 5.5% and the Fitbit Inspire 3 worst at 16.5%. Model matters as much as brand, and irregular arm movement, not intensity, is what breaks wrist readings.

Which smartwatch has the most accurate blood oxygen (SpO2)?

Apple Watch has the most evidence and the smallest errors at normal levels: 2.2% in a low-oxygen lab test (Series 8, Walzel 2023) and 2.2% at rest (Series 7, Jiang 2023). Galaxy Watch 5 measured 3.8% and read low; Garmin Venu 2s measured 5.8% and read low. When oxygen is truly low, watches tend to read high: an Apple Watch Series 7 over-read 85% of readings below 88% against arterial blood (Jiang 2026). Oura and WHOOP have no independent SpO2 validation, and no wearable SpO2 feature is FDA-cleared.

How accurate are wearable calorie counts?

Calories are the least accurate wearable metric. In a 2026 study that tested Apple, Fitbit, Garmin and Samsung on the same 58 people against lab gas analysis, cycling errors ran from 31.7% to 77.6% (Kostrna 2026). Apple, Garmin and Samsung read high on average; Fitbit averaged close to zero, but only after implausible readings were removed. In weight training, Apple, Garmin and Samsung errors ran from 84.5% to 115% (Yun 2026), and a second 2026 study of 62 men found all four brands reading high in both endurance and weight sessions (Lee 2026). No study has tested WHOOP calories or Oura's current calorie algorithm.

Which wearables have FDA-cleared features?

Apple Watch has five FDA-cleared features (ECG, irregular rhythm alerts, AFib History, sleep apnea and hypertension notifications). Samsung has three (ECG, irregular rhythm alerts, sleep apnea). Fitbit has two (ECG, irregular rhythm alerts), Garmin and WHOOP MG one each (ECG). Oura has none. Blood pressure on Samsung, WHOOP, Oura and Pixel Watch is a wellness feature, not FDA-cleared, and no wearable blood oxygen feature is cleared.

Sources

Every claim, traceable.

78 validation studies, 14 FDA records and a few manufacturer documents, each labelled with who funded it. Last updated 13 Oct 2026.

Oura-funded

Accuracy of three commercial wearable devices for sleep tracking in healthy adults

Robbins R, et al. 2024. Sensors, n=35.

Oura-funded

Validity of the Oura Ring Gen 3 against polysomnography over multiple nights in healthy adults

Svensson T, et al. 2024. Sleep Medicine, n=96.

Independent · preprint

Head-to-head comparison of eight consumer sleep trackers against polysomnography

Topalidis P, et al. 2026. Preprint, n=18.

Independent

Performance of seven consumer sleep-tracking devices compared with polysomnography

Chinoy ED, et al. 2021. Sleep, n=34.

Independent

A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults

Miller DJ, et al. 2021. Biosensors, n=6.

Independent

Evaluations of commercial sleep technologies for objective monitoring during routine sleeping conditions

Stone JD, et al. 2020. Nature and Science of Sleep, n=5.

Independent · abstract

Validity of the Galaxy Watch 2 for sleep in patients with obstructive sleep apnea

Byun JH, et al. 2023. Chronobiology International, n=127.

Independent

Validity of two consumer multisport activity tracker and one accelerometer against polysomnography

Budig M, et al. 2022. Sensors, n=49.

Independent

Performance of six consumer sleep trackers vs polysomnography

Schyvens AM, et al. 2025. Sleep Advances, n=62.

Independent

Accuracy of 11 wearable, nearable and airable consumer sleep trackers

Lee T, et al. 2023. JMIR mHealth and uHealth, n=75 (clinical, mean AHI 18).

Independent

Accuracy of wristband Fitbit models in assessing sleep: systematic review and meta-analysis

Haghayegh S, et al. 2019. Journal of Medical Internet Research, n=Meta-analysis.

Independent

Oura Ring Gen 3 in sleep-clinic patients

Herberger S, et al. 2025. Scientific Reports, n=45.

Independent

Oura Ring sleep measurement: systematic review and meta-analysis

Khan S, et al. 2025. OTO Open, n=6 studies, 388 people.

Independent

The validity of Apple Watch Series 9 and Ultra 2 for serial measurements of heart rate variability and resting heart rate

O'Grady B, Lambe R, Baldwin M, Acheson T, Doherty C 2024. Sensors, n=39.

Independent

Agreement between Fitbit Versa 4 and a reference standard on time-domain nocturnal heart rate variability: a pilot study

Nasr N, Behbehani K, Wang J, et al. 2025. Digital Health, n=8.

Independent

Accuracy of photoplethysmography-derived pulse rate variability compared with ECG-derived HRV: systematic review and meta-analysis

Xu, et al. 2026. Sensors, n=43 studies.

Independent

Assessment of Fitbit Charge 4 for sleep stage and heart rate monitoring against polysomnography and during home monitoring in Huntington's disease

Doheny EP, Renerts K, Braun A, et al. 2024. Journal of Clinical Sleep Medicine, n=10.

Independent

Validation of nocturnal resting heart rate and HRV in consumer wearables

Dial MB, et al. 2025. Physiological Reports, n=13 (536 nights).

Independent

Six wearables vs polysomnography and ECG for sleep, heart rate and HRV

Miller, et al. 2022. Sensors, n=53.

Independent

Ten-device wrist heart rate validation

Gielen J, et al. 2026. JMIR Formative Research, n=45 (10 sessions per device).

Independent

Fitbit Charge 4 heart rate across real training sessions and sports

Ceugniez M, et al. 2025. JMIR mHealth and uHealth, n=26.

Independent

Accuracy of Fitbit Charge 2 heart rate during sleep

Haghayegh S, et al. 2019. Chronobiology International, n=35.

Independent

Validation of Fitbit Charge 2 sleep and heart rate in shift workers

Stucky B, et al. 2021. Journal of Medical Internet Research, n=59.

Independent

Five wearables vs 12-lead ECG during exercise

Van Oost L, et al. 2025. Sensors, n=24.

Independent

A comprehensive accuracy assessment of Samsung smartwatch heart rate and heart rate variability

Sarhaddi F, et al. 2022. PLOS ONE, n=28.

Independent

Assessment of physiological signals from photoplethysmography sensors compared to an electrocardiogram sensor in daily life

Zia Ur Rehman R, et al. 2024. Sensors, n=25.

Independent

Apple Watch 7 and Galaxy Watch 4 vs 12-lead ECG during cardiopulmonary exercise testing

Kim J, et al. 2023. Annals of Rehabilitation Medicine, n=44 cardiac patients.

Google-funded

Validity of the Google Pixel Watch 3 for estimating VO2 max and heart rate against a maximal treadmill test

Lambe R, et al. 2026. PLOS ONE, n=26.

Independent

The accuracy of Apple Watch measurements: a living systematic review and meta-analysis

Lambe R, et al. 2026. npj Digital Medicine, n=82 studies.

Independent

Accuracy and precision of energy expenditure, heart rate and steps measured by Fitbit

Chevance G, et al. 2022. JMIR mHealth and uHealth, n=32 heart rate studies.

Independent

Evaluation of leading smartwatches for the detection of hypoxemia

Walzel S, et al. 2023. Sensors, n=18.

Independent

Accuracy of blood oxygen saturation in common consumer smartwatches

Jiang Y, et al. 2023. PLOS Digital Health, n=49.

Independent

Wearable pulse oximetry during controlled hypoxia induction

Jiang Y, et al. 2026. JMIR Formative Research, n=9 (308 blood samples).

Samsung-funded

Smartphone and smartwatch pulse oximetry during hypoxia laboratory testing

Browne SH, et al. 2025. Sensors, n=12 in the watch arm.

Independent

Smartwatch vs polysomnography for overnight oximetry and sleep apnea evaluation

Browne SH, et al. 2024. Journal of Clinical Sleep Medicine, n=51.

Independent

Garmin Fenix 5X Plus oxygen saturation at 4,559 m

Schiefer LM, et al. 2021. Sensors, n=13.

Independent

Consumer wearable clinical parameters in COPD and healthy controls

Hermans F, et al. 2024. JMIR mHealth and uHealth, n=38.

Independent · preprint

Garmin Venu heart rate, respiratory rate and SpO2 in a respiratory ward

Knevels R, et al. 2026. Research Square (preprint), n=29.

Independent

Arterial oxygen saturation from wearables under varying conditions

Hearn, et al. 2023. Aerospace Medicine and Human Performance, n=10.

Independent

Fitbit Sense 2 in hospitalized general medicine patients

Lee, et al. 2026. International Journal of Medical Informatics, n=33.

Apple document

Blood Oxygen app on Apple Watch (white paper)

Apple 2022. Apple, n=50.

Independent

Apple Watch 6 vs Galaxy Watch 4: a validity study of step-count estimation in daily activities

Hong KR, Hwang IW, Kim HJ, Yang SH, Lee JM 2024. Sensors, n=104.

Independent

How valid are wearable physical activity trackers for measuring steps?

An HS, Jones GC, Kang SK, Welk GJ, Lee JM 2017. European Journal of Sport Science, n=35.

Independent

Step count reliability and validity of five wearable technology devices while walking and jogging in a free motion setting and on a treadmill

Montes J, Tandy R, Young J, Lee SP, Navalta JW 2020. International Journal of Exercise Science, n=40.

Independent

The validity of activity trackers is affected by walking speed: Garmin VivosmartHR and StepWatch3

Svarre FR, Jensen MM, Nielsen J, Villumsen M 2020. PeerJ, n=30.

Independent

Comparison of Oura smart ring against ActiGraph accelerometer for physical activity and sedentary time in a free-living context

Niela-Vilén H, Azimi I, Suorsa K, et al. 2022. Computers, Informatics, Nursing, n=42.

Independent

Consumer wearables vs a research-grade monitor for physical activity in free living

Miwa T, et al. 2026. PLOS ONE, n=36.

Independent

Apple Watch step counting during slow and shuffling gait

Rowe G, et al. 2025. PLOS ONE.

Independent

Garmin Vivoactive 4 heart rate, calories and steps on a treadmill

de Leon MF, et al. 2026. Applied Sciences, n=18.

Independent

Four smartwatches in laboratory and field conditions

Rider BC, et al. 2025. Journal for the Measurement of Physical Behaviour, n=12.

Independent

Fitbit Inspire 2 treadmill step accuracy

Cheung YS, et al. 2025. Physiotherapy Practice and Research, n=30.

Independent

Validity of wearable monitors and smartphone applications for measuring steps in semi-structured and free-living settings

Giurgiu M, et al. 2023. Technologies.

Independent

Why your smartwatch may be misleading your doctor: mobility aids and wearable accuracy in older adults

Özel A, et al. 2026. PeerJ, n=42 adults aged 51 to 80.

Independent

Validation of Oura ring energy expenditure and steps in laboratory and free-living (correction 10.1186/s12874-023-02029-w)

Kristiansson E, Fridolfsson J, Arvidsson D, Holmäng A, Börjesson M, Andersson-Hall U 2023. BMC Medical Research Methodology.

Independent

Four smartwatches vs indirect calorimetry during cycling

Kostrna, et al. 2026. PLOS ONE, n=58.

Independent

Four smartwatches during endurance and resistance exercise vs indirect calorimetry

Lee TH, et al. 2026. Sensors, n=62.

Independent

Four smartwatches during resistance exercise

Yun, Ko, Lee 2026. Bioengineering, n=70.

Independent · abstract

Five wrist wearables with sex-specific energy expenditure algorithms

Perez, Davis, Wong, Navalta 2025. Int J Exercise Science Conference Proceedings (abstract), n=97 (about 20 per device).

Samsung-funded

Galaxy Watch 6 and 7 during intermittent running

Ferreira, et al. 2026. JMIR Formative Research, n=148.

Independent

Oura Gen 2 and Polar Vantage vs research monitors in free living

Henriksen A, et al. 2022. JMIR Formative Research.

Independent

Fitbit Alta HR vs doubly labelled water

Shook RP, et al. 2022. Journal of Nutrition, n=24.

Apple document

Heart Rate, Calorimetry, and Activity on Apple Watch (white paper)

Apple 2024. Apple.

Independent

Accuracy in wrist-worn devices for heart rate and energy expenditure

Shcherbina A, et al. 2017. Journal of Personalized Medicine, n=60.

Independent

Validation of aerobic capacity (VO2max) and pulse oximetry in wearable technology

Carrier B, et al. 2025. Sensors, n=19.

Independent

Garmin Fenix 6 with chest strap in an athletic sample

Carrier B, et al. 2023. Technologies, n=20.

Independent

Forerunner 245 VO2 max in highly vs moderately trained runners

Engel FA, et al. 2026. European Journal of Applied Physiology, n=35.

Independent

Apple Watch VO2 max estimation

Caserman P, et al. 2024. JMIR Biomedical Engineering.

Independent

Validation of Apple Watch VO2 max estimates

Lambe RF, et al. 2025. PLOS ONE.

Independent

Accuracy of VO2 max estimates from Apple Watch Series 10

Lambe R, et al. 2026. Mayo Clinic Proceedings: Digital Health, n=40.

Independent

Validity of cardiorespiratory fitness measured with Fitbit compared to VO2max

Klepin K, et al. 2019. Medicine & Science in Sports & Exercise, n=60.

Independent

Fitbit Charge 2 VO2 max estimation

Freeberg KA, et al. 2019. mHealth.

Independent

Validity of consumer wearable VO2 max estimates: meta-analysis

Molina-Garcia P, et al. (INTERLIVE) 2022. Sports Medicine, n=Meta-analysis.

Oura document

How accurate is Oura's temperature data? (internal study, not peer reviewed)

Oura 2020. Oura blog, n=16 people + lab bath test.

Oura-funded

Oura Ring as a tool for ovulation detection

Thigpen N, et al. 2025. Journal of Medical Internet Research, n=964 users, 1,155 cycles.

Apple-funded

Wrist temperature algorithms for ovulation and next-period prediction

Wang Y, et al. 2025. Human Reproduction, n=260 people, 889 cycles.

Samsung-funded

Galaxy Watch 5 wrist temperature for ovulation prediction

Kim SE, et al. 2026. Biosensors and Bioelectronics, n=85 women, 225 cycles.

Samsung-funded

Validating a consumer smartwatch for nocturnal respiratory rate

Jung H, et al. 2023. Sensors, n=195 sleep-clinic patients.

Fitbit-funded

Measurement of respiratory rate using wearable devices

Natarajan A, et al. 2021. npj Digital Medicine, n=28 people, 52 nights.

WHOOP-funded

Effect of wearables on sleep in healthy individuals: validation study

Berryhill S, et al. 2020. Journal of Clinical Sleep Medicine, n=32.

Oura document

How accurate is Oura's respiratory rate? (not peer reviewed)

Oura 2020. Oura blog, n=43.

Independent

Continuous vital sign monitoring in post-operative patients with consumer fitness trackers

Helmer P, et al. 2024. Digital Health, n=12 in the Garmin arm.

Independent

Consumer wearables for continuous monitoring during endoscopy

AbiMansour JP, et al. 2024. Frontiers in Digital Health, n=292 procedures.

FDA record

Apple Hypertension Notification Feature

FDA 2025. 510(k), K250507.

FDA record

Apple Sleep Apnea Notification Feature

FDA 2024. 510(k), K240929.

FDA record

Apple Irregular Rhythm Notification Feature

FDA 2018. De Novo, DEN180042.

FDA record

Apple ECG App

FDA 2018. De Novo, DEN180044.

FDA record

Apple AFib History Feature

FDA 2022. 510(k), K213971.

FDA record

Samsung ECG Monitor App

FDA 2020. 510(k), K201168.

FDA record

Samsung Irregular Heart Rhythm Notification

FDA 2023. 510(k), K230292.

FDA record

Samsung Sleep Apnea Feature (first De Novo)

FDA 2024. De Novo, DEN230041.

FDA record

Samsung Sleep Apnea Feature v2.0 (Galaxy Watch and Galaxy Ring)

FDA 2026. 510(k), K261011.

FDA record

Fitbit ECG App

FDA 2020. 510(k), K200948.

FDA record

Fitbit Irregular Rhythm Notifications

FDA 2022. 510(k), K212372.

FDA record

Pixel Watch Loss of Pulse Detection

FDA 2025. 510(k), K242967.

FDA record

Garmin ECG App

FDA 2023. 510(k), K221774.

FDA record

WHOOP MG Heart Screener (ECG)

FDA 2025. 510(k), K243236.

Show all 96 sources

From the blog

Show fewer sources

Keep reading.

The long-form, evidence-based articles behind this tool.

Wearables & Data

What's the Most Accurate Wearable? 17 Studies, 6 Devices, Ranked (2026)

Seventeen independent studies on sleep, HRV, heart rate and step accuracy, with the actual numbers behind each device.

10 min read

Read

Wearables & Data

How Accurate Is Your Heart Rate Monitor? 10 Devices Ranked

Ten wearables, the same 45 wrists, the same day, and error running from 5.5% to 16.5%. Which device you own matters less than you think.

12 min read

Read

Sleep

Most Accurate Sleep Tracker? Oura vs Apple vs Fitbit vs Garmin vs Whoop (2026)

Every tracker calls itself accurate. Measured against lab polysomnography, the ranking changes completely on one thing most reviews ignore.

8 min read

Read

Kygo Health

Stop Guessing. Start Knowing.

Home How It Works Blog Contact Privacy Terms

This content is for informational purposes only and is not medical advice. Wearable readings are estimates, not clinical measurements. Consult a qualified healthcare provider before making health decisions based on wearable data.

Data from peer-reviewed validation studies, preprints, FDA records and manufacturer documentation. Last updated 13 Oct 2026.

As an Amazon Associate, Kygo Health earns from qualifying purchases. Product links on this page are affiliate links; we may earn a commission at no extra cost to you.

Which wearable is actually accurate?

Compare wearables head-to-head.

FDA-cleared features

Breathing rate

Skin temperature

VO2 max

Calories burned

Step counting

SpO2

Heart rate during activity

Resting heart rate

Overnight HRV

Sleep timing

Sleep stages

Every metric, side by side

No device wins everywhere. Kygo reads them all.

Each brand wins something, none wins everything.

What no brand can fix.

Who wears it beats which brand

Calories are the weakest number

Who paid for the study matters

The newest models are unvalidated

SpO2 is least reliable when it matters

Cleared does not mean it catches everything

Ten metrics plus FDA, one section each.

Sleep stages: Oura leads; who you are matters more.

Sleep timing: Oura, then Fitbit, then Apple.

Overnight HRV: Oura leads.

Resting heart rate: Oura, then WHOOP.

Heart rate during activity: Apple edges Garmin and Fitbit.

SpO2: least reliable when it counts.

Step counting: Garmin leads in daily life.

Calories burned: Fitbit is least bad; none is reliable.

VO2 max: Garmin leads, with a chest strap.

Skin temperature: no independent sensor test.

Breathing rate: only two brands qualify.

FDA-cleared features: cleared is not caught.

Related tools.

Common questions.

Every claim, traceable.

Keep reading.

KYGO

About this wearable accuracy comparison