Metric
Sleep
What your sleep score is built from, why it moves, and what the stages and totals actually mean.
Sleep scores feel precise, but they're built from estimates, and they don't always match how rested you feel. A good night can score badly and a short night can score well. Here's what the research says these numbers are really tracking, what moves them, and how much weight to put on the score versus the trend.
What your device actually calls this
Scientific measurementSleep architecture (light/deep/REM stages) via actigraphy (movement) and/or heart rate-derived signals, not EEG, which is what defines "real" sleep staging in a lab.
| Device | What they call it | What it actually measures |
|---|---|---|
| Oura | "Sleep Score" | 7 inputs: total sleep time, efficiency, restfulness, REM duration, deep sleep duration, latency, timing consistency. |
| Whoop | "Sleep Performance" | 4 inputs: sleep sufficiency (hours vs. needed), sleep consistency (vs. prior 4 nights), sleep efficiency, sleep stress. |
| Garmin | "Sleep Score" | Incorporates a separate stress metric derived from HRV, alongside sleep stages. |
| Fitbit | "Sleep Score" | Combines heart rate, total time asleep, and sleep-stage breakdown into one 0-100 number, grouped into four bands from Poor to Excellent. Available on the free tier. |
| Apple Watch | "Sleep Stages" and, since watchOS 26, "Sleep Score" | Since watchOS 9, estimates Awake/REM/Core/Deep stages from motion, heart-rate dynamics, and circadian timing. watchOS 26 added a 0-100 Sleep Score weighted across sleep duration (50 points), bedtime consistency (30 points), and interruptions (20 points). |
| RingConn | Sleep stages, breathing rate, skin temperature | A real published validation exists: a joint RingConn/Shanghai Ruijin Hospital dataset compared the ring against polysomnography, the clinical gold standard. |
| Samsung Galaxy Watch | "Sleep Score" and "Sleep Coaching" | Sleep Score runs 1-100 from four inputs; Sleep Coaching layers HRV, SpO2, and respiratory rate over 4-5 weeks into a personalized plan. |
| Ultrahuman Ring | "Sleep Algorithm 2.0" | Machine-learning model trained on 125,000+ nights. Ultrahuman's own 6-participant pilot study reported 2.4-2.6% mean absolute error against an FDA-approved reference device, but it's self-published, not independently peer-reviewed. |
| Coros | "Sleep Score" and "Sleep Stress" | Stages graded against National Sleep Foundation thresholds; Sleep Stress separately flags physical/mental strain from elevated heart rate and lowered overnight HRV. |
| Amazfit | No standalone Sleep Score | Sleep tracking feeds directly into the next morning's Readiness Score rather than standing alone as its own branded score. |
| Polar Vantage | "Sleep Plus" | Tracks stages, continuity, and quality, feeding results into Recovery Pro alongside HRV and training data. |
Oura, Whoop, Garmin, Fitbit, and, since watchOS 26, Apple all call it a "Sleep Score" or close to it, but the actual formulas weigh meaningfully different things. Oura leans on stage architecture, Whoop on performance and consistency framing, Garmin folds in a stress signal the others do not, Fitbit runs the simplest formula of the group, and Apple is the only one that publishes its exact weighting. A Sleep Score from one device is not measuring the same combination of things as another, even at the same numeric value.
Device measurement accuracy: Uncertain · on the Research DeskHow each device actually shows this
Oura
"Sleep Score"
Whoop
"Sleep Performance"
Garmin
"Sleep Score"
Fitbit
"Sleep Score"
Apple Watch
"Sleep Stages" and, since watchOS 26, "Sleep Score"
RingConn
Sleep stages, breathing rate, skin temperature
Samsung Galaxy Watch
"Sleep Score" and "Sleep Coaching"
Ultrahuman Ring
"Sleep Algorithm 2.0"
Coros
"Sleep Score" and "Sleep Stress"
Amazfit
No standalone Sleep Score
Polar Vantage
"Sleep Plus"
See every device →
Sources: Oura's own support docs (support.ouraring.com) and Whoop's own published sleep-accuracy page (whoop.com), both primary. Fitbit/Google's own support documentation (official). Apple's own support documentation (official). Garmin's stress-in-sleep-score detail came from a secondary comparison source.
The benchmark
Sleep
Sleep Score Explained: What Is a Good Sleep Score?
What goes into the number and what it can and can't tell you.
See how you compareExample result
Age 40 ยท 6.5 h
just below the recommended range
NSF expert consensus
More on Sleep
Labs Score REM From Brain Waves, Wearables Estimate It
Labs score REM from brain waves, eye movements and muscle tone. Versus a lab, Oura missed REM by 4 to 6 minutes and WHOOP by 21 in reviewed studies.
Research Sets Total Sleep Targets, Not a Deep Sleep Number
Sleep targets are for total sleep, 7 to 9 hours. Deep sleep has been linked to AFib odds and brain MRI markers, but no study sets a minimum for it.
Does Poor Sleep Increase Dementia Risk?
Sleep-architecture research ties slow-wave sleep to memory consolidation and neurodegenerative risk, not to your tracker's nightly deep sleep score.
Watches Are Only So-So at Guessing Sleep Stages
Six watches scored sleep stages against a sleep lab and matched only fairly (kappa 0.21 to 0.53). Sleep versus wake was far more accurate.
Six Nights at Four Hours Produced Changes Like Aging
Six nights at four hours in bed produced hormone and glucose changes researchers called similar to aging. Eleven young men, and not proof that sleep ages you.
No Trial Has Followed Insomnia Treatment to a Heart Outcome
No trial has followed people treated for insomnia through to a heart outcome. What exists links disrupted sleep to strain on heart-regulating systems.
What Is Sleep Banking, and Does It Actually Work?
Sleep banking and napping show measurable effects on alertness and stress hormones, but research finds some sleep debt costs don't fully reverse.
How Caffeine Timing Actually Affects Your Sleep
Caffeine close to bedtime cuts total sleep time, delays sleep onset and pushes sleep toward lighter stages. How early you have to stop depends on the dose.
Does a Consistent Sleep Schedule Actually Matter?
Irregular sleep timing tracked with worse outcomes independently of how many hours people slept. Consistency and duration are two different measurements.
REM Sleep Behavior Disorder, Parkinson's Risk, and What Wearables Can Actually Catch
Sleep-lab studies established that REM sleep behavior disorder precedes Parkinson's. The wearable detection work is all preprint and diagnoses nobody.
Is Fitbit Sleep Tracking Accurate? How Garmin and WHOOP Compare
Research comparing Fitbit, Garmin, and WHOOP against sleep lab recordings shows total sleep time tracks better than light, deep, and REM stage detection.
Should You Stop Tracking Your Sleep?
Anxious fixation on a sleep score has been measured in a general population sample, not just described in case reports. Here is what that study found.
Evening Workouts Went With Worse Sleep Only Close to Bed
Exercise ending 4+ hours before sleep left sleep unchanged in 14,689 people. Later, harder sessions went with worse sleep.
How Accurate Are Under-Mattress Sleep Trackers Compared With Wearables?
The Withings Sleep Analyzer overstated sleep by 48 minutes in a sleep lab, catching 95% of sleep but 37% of waking time. How wrist trackers and rings compare.
Deep-Sleep Sounds Boost Brain Waves, Not Clearly Memory
Sound timed to deep sleep raised slow waves, but memory effects were small (g 0.25, p 0.07) across 177 people and shrank over time.
How Long Does It Take to Recover From Sleep Debt?
One recovery night restores how alert you feel. Attention and metabolic markers lag well behind, and no single recovery time has been established.
Can Wearables Detect Deep Sleep Accurately?
Wearables track total sleep time reasonably well, but against polysomnography deep sleep staging is where they consistently struggle most.
Insomnia's Sleep Pattern Looks Like a Milder Depression
A meta-analysis of sleep-lab studies found insomnia, depression and narcolepsy fall on one continuum of worsening sleep disturbance, not distinct patterns.
REM-Stage Sleep Apnea Tracked With Memory More Than Total Severity
In 81 adults at elevated Alzheimer's risk, REM-stage sleep apnea features tracked with verbal memory more closely than the standard whole-night score.
Sleep Banking Before a Rough Stretch Showed a Real Benefit
Two controlled studies found extending sleep for a week before deprivation preserved performance during it and sped up recovery afterward.
What Is the Most Accurate Sleep Tracker?
In head-to-head sleep-lab tests, the Oura Ring matched the lab most closely on sleep stages and the Apple Watch led the wrist devices. Every tracker is closer on total sleep than on stages.
Is REM Sleep Really for Memory and Recovery, or Something Else?
REM is usually explained as your brain's nightly repair and memory shift. A rival, evidence-backed theory says it's doing something else entirely.
Your Tracker Says You Slept Badly. You Feel Fine. Who's Right?
A low sleep score next to a morning that feels normal isn't a glitch. Research on self-reported sleep shows the two signals are built to diverge.
Does Deep Sleep Actually Drive Your Testosterone?
Hormone-optimization advice treats deep sleep as a testosterone lever. A controlled trial that suppressed only deep sleep tested that link directly.
Is Your Deep-Sleep Number Connected to Your Blood Pressure?
Insomniacs staring at a low deep-sleep score often wonder if it's hurting their heart. A 1,607-person study tested that exact link.
Does Pollen Season Actually Wreck Your Sleep?
A two-week study tracked pollen exposure against wearable-measured sleep in 373 adults to test the common assumption that allergy season ruins sleep.
Does More REM Sleep Mean You're Less Anxious, or Is It the Other Way Around?
A study measuring trait anxiety against wearable-tracked sleep stages found a relationship that runs opposite to what most people would guess.
Older Adults Sleep Less, but 7 to 8 Hours Still Fits Health
Measured sleep falls 10 minutes per decade of age. The sleep amount linked to the best health stayed 7 to 8 hours in adults.
Why Your Tracker Might Show Zero or Near-Zero Deep Sleep on an Otherwise Normal Night
A validation study found an Apple Watch underestimated deep sleep by 43 minutes and a Fitbit by 15, while an Oura Ring matched lab monitoring.
Why Your REM Sleep Number Might Read Low, and Why the Device Is Less of a Culprit Here
A 96-participant study found the Oura Ring underestimates REM sleep by 4.1 to 5.6 minutes, but its 90.6 percent REM accuracy is its best stage.
Does Ozempic Help Sleep Apnea?
The phase 3 trial that cut apnea events by 23.8 an hour tested tirzepatide, not semaglutide. Only tirzepatide carries an FDA approval for sleep apnea.
Does Intermittent Fasting Improve Sleep?
Across 16 randomized trials, intermittent fasting produced real weight, fat mass and blood pressure reductions and no overall effect on sleep quality.
Is Oura Sleep Tracking Accurate? What the Lawsuit's Numbers Actually Measure
The Oura lawsuit pits a 95% accuracy claim against a 53.18% study result. Both figures are real. They measure different things in different people, and each traces to a specific published study.
What's the Difference Between Core Sleep and Deep Sleep?
Core is Apple Watch's name for stages N1 and N2, what other trackers call light sleep; Deep is stage N3. Why a night that is mostly Core is normal, and why the watch's most common error is scoring deep sleep as Core.