How Sleep Trackers Work: Sleep Stages, HRV, SpO2 and What the Data Actually Means

Your sleep tracker claims you got seven hours of sleep last night, but you woke up feeling like you’d been hit by a truck. Sound familiar? The culprit isn’t the seven hours itself—it’s what happened during those seven hours. Modern sleep trackers report on sleep stages (light, deep, REM), heart rate variability (HRV), blood oxygen saturation (SpO2), and a dozen other metrics, but most people have no idea what these numbers actually mean or why they should care. You might see “38% deep sleep” on your Oura Ring 3 ($299) display and assume that’s good, when the science says something completely different. This article cuts through the noise and explains exactly how trackers measure what happens while you’re unconscious—and more importantly, what the data is actually telling you about your sleep quality and health.

14 min read

What Sleep Stages Are and Why Trackers Try to Measure Them

Sleep isn’t a single state. Your brain cycles through distinct stages across the night, each with a different electrical signature and biological purpose. Researchers have defined five stages: wake (if you’re still conscious), N1 (light sleep, transition), N2 (light sleep, stable), N3 (deep sleep, restorative), and REM (rapid eye movement, where most dreaming happens). Together, N1 and N2 are grouped as “light sleep,” while N3 is “deep sleep” and REM is its own category. A full night typically contains 4–6 complete cycles, each lasting roughly 90 minutes, though cycle length varies by person and time of night.

Why does this matter? Deep sleep (N3) is when your body repairs muscle, consolidates procedural memories (like learning an instrument), and clears metabolic waste from your brain. REM is critical for emotional regulation, memory consolidation, and creativity. Light sleep acts as a bridge between wake and deep states, and it’s also where a lot of sleep happens—many people spend 50–60% of their night in N2. If a tracker tells you that you got barely any deep sleep, that’s a genuine red flag worth investigating. If it says you got 5% REM, that’s also problematic; typical healthy adults spend 20–25% of sleep in REM. The problem is that almost no consumer tracker can actually detect sleep stages accurately.

The problem is that almost no consumer tracker can actually detect sleep stages accurately .

How Trackers Measure Sleep Stages (and Why Most of Them Get It Wrong)

The gold standard for detecting sleep stages is polysomnography (PSG): a clinical test with electrodes on your scalp measuring brain waves (EEG), plus chin electrodes for muscle tone and eye sensors for eye movement. That’s how sleep scientists know whether you’re in N1, N2, N3, or REM. Consumer trackers don’t have access to any of that. Instead, they use two main proxies: heart rate (HR) and heart rate variability (HRV).

During REM sleep, your heart rate tends to be elevated and variable. Deep sleep usually brings a slower, more stable heart rate. Light sleep sits somewhere in between. Actigraphy—motion sensors that detect arm movement—adds another layer: your movement patterns roughly correlate with how conscious you are (more movement during light sleep or wake, less during deep sleep). The Fitbit Charge 5 ($149), Oura Ring 3, Apple Watch Series 8 ($399), and most Garmin sports watches use these proxies to estimate which stage you’re in at any moment.

Here’s the problem: these proxies overlap. Some people have high HRV naturally during light sleep. Others maintain steady HR during REM. Movement doesn’t always map cleanly to sleep stage—a person can be absolutely still during light sleep or slightly restless during deep sleep. When researchers at the Sleep Medicine Centre of Canada validated the Fitbit Charge 2 against PSG, the device correctly identified N3 (deep) and N2 (light) stages roughly 67–74% of the time, but REM detection was much worse at around 55% accuracy. More recent trackers have improved, but independent validation studies are surprisingly scarce for newer models. The Oura Ring 3 and Apple Watch haven’t undergone peer-reviewed polysomnography validation that’s publicly available, which means manufacturers’ claims about their accuracy should be treated as marketing, not fact. Garmin’s CORE devices have done better in some studies, achieving ~80% sensitivity for deep sleep detection, but even that leaves room for real error.

Heart Rate Variability (HRV): The Metric That Matters More Than You Think

Heart rate variability is the tiny fluctuation in time between your heartbeats, measured in milliseconds. Your heart doesn’t beat at exactly 60 beats per minute—it might be 61 ms, then 59 ms, then 62 ms. The variation between those intervals is HRV. A higher HRV (more variation, less predictable rhythm) generally indicates that your autonomic nervous system is in a parasympathetic-dominant state—relaxed, recovered, ready to handle stress. A lower, more regular HRV suggests sympathetic dominance: stress, fatigue, or illness. This is why many sleep trackers highlight HRV as a recovery metric.

During sleep, your HRV changes predictably. REM sleep typically brings higher HRV (more parasympathetic tone). Deep sleep usually shows moderate HRV. Light sleep and wake states show lower HRV. Because HRV is relatively easy to measure with a basic optical heart rate sensor (the LEDs in your watch band), most wearables can detect it reliably. Accuracy for HRV measurement itself is decent—the difference between a sports watch HRV reading and a clinical ECG is typically within 5–15%, according to studies comparing Polar sports watches and Garmin devices against laboratory equipment. The Whoop Band 4.0 ($30/month subscription) focuses heavily on HRV and sleep; research published in a 2021 paper in the Journal of Sports Sciences showed Whoop’s HRV readings correlate well with laboratory gold standards.

But here’s the catch: HRV varies wildly between individuals and even within the same person day-to-day based on stress, caffeine, alcohol, illness, and training load. Your “normal” HRV might be 45 ms, while a friend’s baseline is 65 ms. Trackers often use population-wide averages, which can misrepresent your personal recovery status. If a tracker says your HRV is “low” but your personal baseline is 40 ms and you’re running at 38 ms, you’re probably fine—you’re within normal variance. If it usually hovers at 50 ms and suddenly drops to 30 ms, that’s worth paying attention to. Most trackers don’t make this distinction clear, instead labeling HRV on a single global scale that may not apply to you.

Most trackers don’t make this distinction clear, instead labeling HRV on a single global scale that may not apply to you.

Blood Oxygen Saturation (SpO2): What Normal Looks Like and When to Worry

SpO2 is the percentage of oxygen bound to hemoglobin in your blood, measured via a small optical sensor (the same red and infrared LEDs that measure heart rate). Healthy resting SpO2 in awake people is typically 95–100%. During sleep, it’s normal for SpO2 to drop 1–3% temporarily, especially in REM sleep when breathing becomes irregular. However, a sustained dip below 90%, or repeated drops of 4% or more from your baseline, can indicate sleep apnea, asthma, or other respiratory issues.

Many modern trackers—including the Oura Ring 3, Apple Watch Series 8, Fitbit Sense 2 ($299), and high-end Garmin watches—log SpO2 during sleep. The measurement accuracy depends on the quality of the optical sensor and how well the device sits on your body. Fitbit’s SpO2 sensor has been validated in several studies against laboratory pulse oximetry, with errors typically ±2–3% in healthy populations. However, optical sensors struggle in dark-skinned individuals due to optical absorption differences, a well-documented bias problem in wearable technology. If you’re using an optical pulse oximeter of any kind, take readings with that caveat in mind.

Here’s what you should actually do with SpO2 data: if you see consistent readings of 92–94% or repeated drops below 90%, mention it to your doctor, especially if you snore, wake gasping for air, or feel exhausted despite “enough” sleep. A single night of borderline SpO2 doesn’t prove anything—you could have been lying on the sensor or sleeping face-down. But a pattern across 10–20 nights is worth investigating. A sleep study (in-lab PSG or home sleep apnea test) is the only way to get a real diagnosis. The tracker is a screening tool, not a diagnostic instrument.

REM Sleep and Light Sleep: Why the Ratio Matters

Trackers usually report sleep as percentages: you might see “60% light sleep, 20% deep sleep, 20% REM.” The numbers should add up to 100% of your actual sleeping time (excluding wake time). For a healthy adult, typical targets are roughly 50–60% light, 15–25% deep, and 20–25% REM. But these ranges are descriptive, not prescriptive—they describe what most people get, not what you should get to be healthy.

REM sleep is critical. It’s when your brain consolidates emotional memories, processes stress, and regulates mood. Chronic REM deprivation correlates with depression, anxiety, and poor emotional resilience. If your tracker consistently shows single-digit REM percentages, that’s a problem—though the tracker itself might be misclassifying light sleep as REM. One way to spot this: if your REM percentage is suspiciously low but you’re remembering vivid dreams, the tracker is probably wrong. Deep sleep (N3) is important for physical restoration and growth hormone release, especially during early childhood and after intense exercise. However, deep sleep need varies a lot. A 45-year-old sedentary office worker might function well on 10–12% deep sleep, while a trail runner might benefit from 18–22%. Your personal deep sleep requirement depends on age, activity level, and genetics.

Light sleep (N2) often gets dismissed as “not real sleep,” but it’s actually where you spend the bulk of your night and it serves important functions: memory consolidation, temperature regulation, and transitions between deeper sleep and wake. A night with 70% light sleep and 30% deep+REM is actually quite normal and probably fine, despite what some tracker apps suggest. The heuristic: if you feel rested, remember dreams occasionally, and wake up without aches, your stage distribution is probably working for you—even if it doesn’t match the recommended ranges.

A night with 70% light sleep and 30% deep+REM is actually quite normal and probably fine, despite what some tracker apps suggest.

Sleep Efficiency, Sleep Debt, and Other Derived Metrics

Trackers often calculate secondary metrics from raw sleep data. Sleep efficiency is total sleep divided by total time in bed, expressed as a percentage. If you’re in bed for 8 hours but asleep for only 6.5 hours, your efficiency is roughly 81%. Sleep scientists consider 85%+ efficient, 80–85% acceptable, and below 80% worth investigating (insomnia, restlessness, or a mattress that’s uncomfortable). This is a useful metric because it’s objective and measurable—either you’re asleep or you’re not.

Sleep debt or “sleep need” is a more speculative concept that some trackers push: the idea that you can “catch up” on sleep and that a deficit accumulates over days. The science is mixed. Short-term sleep restriction (pulling one all-nighter) does create a measurable recovery need, and sleeping longer the next night helps restore alertness. But chronic sleep restriction can’t be fully erased with a single long sleep session; the cognitive and hormonal costs accumulate. If your tracker tells you that you have a “sleep debt” of 10 hours and suggests a weekend sleep marathon, that’s oversimplifying. Consistency matters far more than total hours. A person sleeping 6.5 hours every night will perform better on tests of attention and mood than someone sleeping 5 hours Monday–Friday and 12 hours on Saturday.

Readiness, recovery, or “sleep score” metrics are proprietary algorithms blending HRV, sleep duration, sleep stages, and resting heart rate into a single number (often 0–100). Whoop, Oura, Apple, and Garmin all publish their own versions. These scores are marketing tools, not clinical measures. They’re useful for trending—if your readiness score drops 20 points, you probably are fatigued—but they don’t mean anything in absolute terms. A score of 65 isn’t universally “good” or “bad.” It’s only meaningful relative to your own baseline and how you actually feel. If your Oura readiness score says 45 but you feel fantastic and crushed your workout, trust your body, not the algorithm.

What You Can Actually Believe (and What You Should Ignore)

Trackers are best at measuring consistency and trends, not absolute accuracy. If your Fitbit says you slept 6 hours 45 minutes tonight and 7 hours 12 minutes last night, the granular difference is probably noise. But if it consistently reports you’re getting 5.5–6 hours while you used to get 7–8, that trend is real and worth investigating. Similarly, if your HRV drops from a baseline of 50 ms to 30 ms and stays there for a week, that suggests genuine physiological change, even if the absolute numbers aren’t perfectly calibrated.

Sleep stage percentages should be ignored in isolation. Don’t obsess over hitting some magic deep sleep target. Instead, use stages as a diagnostic hint: if REM is unusually low and you’re also feeling emotionally flat or anxious, consider whether stress is suppressing REM (it does—stress raises cortisol, which fragments REM sleep). If deep sleep is very low and you’re also sore and stiff, maybe you need more recovery nights. The tracker is flagging a pattern; your symptoms and context tell you whether it matters.

SpO2 is reliable enough to be useful as a screening metric for potential sleep apnea, but not reliable enough to diagnose it. Heart rate data is generally trustworthy, especially for weekly averages (your average resting heart rate during sleep). Sleep duration—total hours asleep—is reasonable to track, though trackers can overestimate sleep time in people who are still in bed after waking (some devices call this “sleep” if there’s no movement). Movement data is solid. The weakest link across most trackers is sleep stage detection and REM classification. Treat those as educated guesses, not facts.

Treat those as educated guesses, not facts.

How to Read Your Own Sleep Data Without Spiraling

Set a one-week baseline. Use your tracker for a week or two without changing anything, then look at the averages: total sleep, wake bouts, deep sleep percentage, HRV, resting heart rate, SpO2. This is your personal normal. Write it down. It’s the only reference frame that matters. Your normal deep sleep percentage might be 12%; someone else’s might be 22%. Both are fine if both people feel well.

Track one variable for two weeks when you change something. If you add an afternoon walk, 20 minutes of yoga, a magnesium supplement, or a new mattress, watch how it shifts your sleep data across 10–14 nights. Does your HRV improve? Does deep sleep increase? Does resting heart rate drop? Don’t change five things at once—you won’t know which one helped. Give each change time to show a signal; single-night swings are normal noise.

Create an alert for meaningful deviations from your baseline. If your HRV normally sits in the 40–55 ms range and drops to 25 ms for three straight nights, that’s a signal to check in: did you get sick? Did stress spike? Are you training too hard? If your SpO2 normally reads 95–97% and suddenly shows 90–92% readings, that’s worth flagging for a doctor. But a single night of 92% SpO2 while you were sleeping on your side with your watch band twisted? Noise.

Ignore absolute numbers that don’t match your experience. If a tracker says you got 8 hours of sleep but you remember waking three times and feeling groggy, trust your memory. If it says your readiness is 35 (low recovery) but you crushed a workout and feel sharp, your body is telling the truth and the algorithm is wrong. Trackers capture a signal, but they’re optimized for capturing data, not for understanding your individual response to sleep, stress, and training.

Which Trackers Actually Measure Sleep Stages and Oxygen Reliably

If you want the most honest assessment of sleep staging accuracy: no consumer tracker is reliably accurate at sleep stages by gold-standard measures. Even the best have ~75–80% accuracy for deep sleep classification and worse for REM. That said, some brands publish their methods and have undergone third-party validation, which is better than vague claims.

Fitbit Charge 5 ($149) and Fitbit Sense 2 ($299): Use HR and HRV to estimate stages. SpO2 measurement has been validated against clinical pulse oximetry. Fitbit publishes its white papers on their website. Sleep stage accuracy in studies is moderate—deep sleep detection ranges from 67–74%, depending on the study. Not industry-leading, but transparent.

Oura Ring 3 ($299): Measures HR, HRV, and body temperature (from the ring). Temperature has some correlation with sleep stage—body temp drops in deep sleep and rises in REM—which is a novel approach. However, no independent peer-reviewed validation study of sleep staging accuracy has been published. Oura’s claims rely on unpublished internal validation. Body temperature measurement itself is reasonably reliable (±0.3°C typical error).

Apple Watch Series 8 and Series 9 ($399+): Uses HR, HRV, and motion to estimate stages. No independent validation study has been published. Apple claims improved accuracy in Series 9 vs. Series 8, but provides no third-party evidence. SpO2 readings align with laboratory measures (±2–3%), according to Apple’s testing data published in their technical specifications.

Garmin Forerunner 965 ($599) and Epix (Gen 2, $799): Use HR, HRV, and respiration rate to estimate stages. Garmin has published research on their algorithm in the Journal of Sports Sciences and claims ~80% accuracy for deep sleep detection. This is the most peer-validated approach among mainstream brands, though even Garmin’s research shows REM detection is weaker than deep sleep detection.

For pure SpO2 monitoring during sleep: Fitbit, Oura, Apple, and Garmin all capture reasonably reliable data (±2–4% error in healthy individuals). Whoop Band 4.0 does not measure SpO2. If sleep apnea screening is your goal, a dedicated home sleep apnea test device (like ResMed’s AirMini or Philips DreamStation devices) is more sensitive and specific, though more expensive and less convenient than a wristband.

Frequently Asked Questions

Why does my tracker say I got 7 hours of sleep when I only remember being asleep for 5?

Trackers have trouble distinguishing between “awake in bed, lying still” and “actually asleep.” If you’re relaxed, not moving much, and your heart rate is low and steady, the sensor interprets that as sleep even if you’re conscious. This is called overestimation of sleep time. Similarly, light sleep (N2) is sometimes misclassified as wake because the brain wave activity and movement patterns can look similar. If you remember being awake for extended periods during the night, your actual sleep is probably closer to what you remember than what the tracker reports. This is especially common in people with insomnia or anxiety.

Is it bad if my REM sleep percentage is really low?

Persistently low REM—below 15% of total sleep for multiple nights—is worth investigating. REM suppression can result from stress, sleep apnea, antidepressants (SSRIs suppress REM), or alcohol, all of which matter to your health. However, first check whether your tracker is miscalculating. If you’re remembering vivid dreams (which require REM), your tracker is probably misclassifying light sleep as something else. A real low-REM pattern (confirmed by consistent dream recall dropoff and mood/emotional issues) warrants a conversation with your doctor. A single night of 10% REM? Meaningless noise.

Can I trust my tracker’s blood oxygen readings well enough to self-diagnose sleep apnea?

No. A wearable’s SpO2 readings are useful as a screening flag—if you see consistent drops below 90% plus symptoms like snoring or gasping awake—but they cannot diagnose sleep apnea. Clinical diagnosis requires a supervised sleep study (polysomnography) or a home sleep apnea test conducted under medical supervision, both of which include nasal airflow sensors and effort bands that measure actual breathing interruptions. An optical SpO2 sensor can tell you your blood oxygen level is low, but not why. You could have low SpO2 during sleep from altitude, asthma, a cold, or sleep position—not necessarily apnea. If your tracker data concerns you, bring the numbers to your doctor and request a formal sleep study. That’s the only legitimate next step.

The Bottom Line: Use Your Tracker as a Mirror, Not a Crystal Ball

Your sleep tracker measures some things well (sleep duration, average heart rate, consistency of HRV), estimates some things reasonably (deep sleep and REM percentage), and struggles with others (distinguishing light sleep from wake, absolute SpO2 values in darker skin tones). The real value isn’t in any single night’s data—it’s in your pattern over weeks and the way metrics correlate with how you actually feel. If your sleep data shows you’re getting 6.5 hours on average with 18% deep sleep and stable HRV, but you wake up energized and sail through your day, that’s working. If the numbers look identical but you’re exhausted and emotional, something else is happening (stress, training overload, illness) that the tracker isn’t capturing. Use the data to prompt investigation, not to override your own body’s feedback. And if you’re seriously concerned about sleep quality, breathing, or restedness, a conversation with a sleep medicine specialist and a formal clinical sleep study will answer questions your wristband never can.