Sleep · Explainer
What Sleep Trackers Actually Measure
Your ring is not reading your brain waves. It is inferring sleep stages from movement and heart rate, and it is better at trends than at last night.
By Julien Malet, · Updated 2026-09-07
Wearable sleep tracking is genuinely useful and routinely over-interpreted. The gap comes from one fact most owners do not know: these devices do not measure sleep. They estimate it.
What the gold standard is
Clinical sleep staging is done with polysomnography — EEG electrodes reading brain activity, plus eye movement, muscle tone, airflow, oxygen saturation and heart rhythm. Sleep stages are defined by brain-wave patterns, so EEG is not one input among several; it is the thing the definition rests on.
A consumer wearable has none of that.
What your device has
Actigraphy — an accelerometer detecting movement. This is the oldest and most reliable signal, and it is genuinely good at one thing: distinguishing sleep from wake. Validated for decades.
Photoplethysmography — a green or infrared LED measuring blood-volume changes at the skin to derive heart rate and heart-rate variability.
Skin temperature, on some devices.
From those, an algorithm infers what your brain is doing. Deep sleep is inferred from low movement plus low heart rate plus higher HRV. REM is inferred from a characteristic pattern of variable heart rate with near-absent movement.
It is educated guessing from peripheral signals, and the algorithms are proprietary and change with firmware updates — which is why your "deep sleep" can shift after an update without your sleep changing at all.
How accurate, specifically
Validation studies against polysomnography have converged on a consistent picture:
Total sleep time: good. Modern wearables are typically within about 20–30 minutes, and often better. Trust this number.
Sleep vs wake: good, though most devices systematically overestimate sleep — they read still-but-awake as asleep. If you lie motionless unable to sleep, your ring may score it as sleep.
Sleep staging: considerably weaker. Agreement with PSG on specific stages is moderate at best, with deep and REM the least reliable. Errors of an hour or more in a night's deep sleep are common.
HRV and resting heart rate: good, and arguably the most valuable outputs. Both are real physiological measurements rather than inferences, and both are sensitive to alcohol, illness and training load.
The score problem
Every device now produces a nightly score, and those scores are algorithmic composites of the least reliable measurements — weighted by choices the manufacturer does not publish.
There is also a documented downside. Clinicians have described orthosomnia: anxiety about sleep driven by tracker data, where the worry about a poor score produces the poor sleep. If your score is making you tense at bedtime, the device is now the problem it claims to solve.
A useful test: does the number tell you anything your own sense of how you slept did not? If not, it is decoration.
What they are genuinely good for
Trends over weeks, not verdicts on last night. Single nights are noisy. A four-week average is informative.
Quantifying what you already suspect. Watching alcohol reliably cut your deep sleep estimate and raise your resting heart rate is more persuasive than being told it does.
Consistency. Bedtime and wake-time regularity are among the more actionable things a tracker shows, and both are real measurements rather than inferences.
Illness and training load. A resting heart rate several beats above baseline is a real signal, often a day before you feel it.
What they cannot do
Diagnose sleep apnoea. Some devices flag breathing disturbances or oxygen dips, and that can reasonably prompt you to get tested — but it is not a diagnosis. Untreated apnoea is common and genuinely serious. Loud snoring plus daytime sleepiness plus witnessed pauses warrants a real sleep study.
Diagnose insomnia, which is a clinical picture, not a graph.
Replace CBT-I. Cognitive behavioural therapy for insomnia is the first-line treatment with the best evidence, better than medication for long-term outcomes, and no wearable substitutes for it.
The devices, briefly
They differ less than the marketing suggests, because they are all inferring from the same two or three signals. Where they differ is form factor and what they do with the data.
Oura Ring — the one I would buy. Not because its algorithm is provably better, but because a ring is the form factor people keep wearing to bed, and adherence is what makes trend data usable. Good HRV and resting heart rate. Requires a subscription on top of the hardware, which is worth knowing before you buy.
Whoop — no screen, subscription-only pricing, built around recovery and strain scores. The most opinionated interpretation layer of the group, which you will either find useful or find is telling you how to feel about your own morning.
Garmin — the best of them if you also train seriously, because sleep is one feature inside a genuinely strong training platform. Sleep staging is no better than the others.
Apple Watch — good total sleep time and now respectable trend reporting, plus an FDA-cleared sleep apnoea notification feature, which is a real differentiator. The obstacle is battery: charging it before bed is how most people end up not wearing it overnight.
Fitbit — the cheapest credible entry point and long-standing sleep tracking. Much of the useful analysis now sits behind Fitbit Premium.
Eight Sleep and other under-mattress trackers — nothing on your body, which solves adherence completely, and they measure movement and heart rate through the mattress rather than at the wrist. Fine for trends, and expensive.
The honest ranking: if you want sleep data specifically and will act on trends, Oura. If you train hard, Garmin. If you already own an Apple Watch, use it and do not buy anything.
If you are buying one
The honest ranking of what matters: total sleep time, resting heart rate, HRV, consistency. If a device reports those and you wear it nightly, the brand matters less than the marketing suggests.
Rings are better tolerated for sleep than watches, which is why adherence tends to be higher — and adherence is what makes trend data usable.
Spend the money if you will act on the trends. If you would only read the score and worry, the free interventions are where the actual sleep is.