How sleep trackers work
Last reviewed: 14 September 2026
Your watch says you got 48 minutes of deep sleep. It did not measure that — it estimated it from how much you moved and what your heart was doing. Here is what that is worth.
Good at trends, shaky on stages
Sleep trackers are reasonably good at telling sleep from wake, and considerably less reliable at breaking sleep into stages. Your deep sleep number is an estimate inferred from movement and heart rate, not a measurement of your brain.
That does not make them useless. It makes them useful for a different thing than most people use them for.
How a tracker decides you are asleep
The clinical standard is polysomnography: an overnight recording in a lab with an EEG measuring brain activity directly, plus eye movement and muscle tone. That is how sleep stages are actually defined, and it is the only way to score them properly.
A watch or ring has none of that. It infers sleep from:
- Movement, via an accelerometer — the oldest signal, and the basis of actigraphy, which has genuine research use.
- Heart rate and, more informatively, heart-rate variability, which shifts characteristically between stages.
- Breathing rate, derived from the same optical sensor.
- Skin temperature on some devices.
An algorithm turns those into a stage estimate. It is a genuinely clever inference and it is still an inference.
So how accurate is it?
The pattern across validation studies is consistent: devices are good at detecting that you are asleep, and much weaker at saying which stage. They also tend to overestimate sleep, because lying very still while awake looks a lot like sleeping.
| What it reports | How much weight to give it |
|---|---|
| Time asleep | Reasonably reliable, though often a slight overestimate. |
| Bedtime and wake time | Reliable, and genuinely the most useful thing it records. |
| Night-to-night trends | Useful. Even an imperfect measure applied consistently shows real direction of travel. |
| Deep / REM / core split | Treat as approximate. Stage-level agreement with lab measurement is moderate at best. |
| Sleep score | A proprietary composite. Every brand computes it differently, so scores are not comparable between devices — or between you and anyone else. |
What a "sleep score" actually is
A sleep score is a number a manufacturer invents by weighting several of its own estimates — duration, continuity, stage proportions, resting heart rate — into a single figure out of 100. There is no standard behind it. An 85 on one device and an 85 on another do not mean the same thing, and neither is a clinical measurement.
Chasing a perfect score is a poor use of attention for a specific reason: anxiety about sleep reliably damages sleep. Clinicians have started using the term orthosomnia for people whose sleep got worse as they tried harder to optimise the numbers on their tracker. If checking the score in the morning changes how you feel about your day, that is a signal to stop checking it.
Ignore the nightly score. Look at the weekly and monthly view of two things only: how long you slept and how consistent your times were. Those are the metrics a tracker measures well, and they are also the two that matter most.
The one thing worth taking seriously
Several devices now flag signs of breathing disturbance during sleep. These are screening signals rather than diagnoses, and they can be wrong in both directions — but a persistent flag, particularly alongside loud snoring, waking with a dry mouth or headache, or daytime sleepiness despite adequate hours, is worth taking to a GP or doctor. Sleep apnoea is common, frequently missed, and very treatable.
That is a genuinely good use of a consumer device: not optimising a score, but noticing a pattern that warrants a conversation with a doctor.
Sources. Self-help advice follows NHS guidance on insomnia, sleep and tiredness, alongside the American Academy of Sleep Medicine's recommendations. Cognitive behavioural therapy for insomnia (CBT-I) is the treatment the NHS recommends for persistent insomnia. This page is general wellbeing information, not medical advice — see a GP or doctor if sleep problems have lasted months or are affecting your daily life.
Common questions
How accurate are sleep trackers?
Reasonably good at detecting whether you are asleep, and considerably weaker at identifying which stage. They also tend to overestimate sleep, because lying still while awake resembles sleeping. Trust the totals and the trends more than the stage breakdown.
How does a sleep tracker know I'm asleep?
It infers it from movement, heart rate and heart-rate variability, breathing rate, and on some devices skin temperature. The clinical standard, polysomnography, measures brain activity directly with an EEG — which is how stages are actually defined.
What is a good sleep score?
There is no standard. A sleep score is a proprietary composite each manufacturer calculates differently, so scores are not comparable between devices. Time asleep and consistency of timing are more meaningful than any score.
Why does my Apple Watch say I got so little deep sleep?
Possibly because you did — deep sleep declines with age and is a smaller share of the night than people expect. But it is also an estimate rather than a measurement, and stage-level accuracy is the weakest part of any consumer tracker.
Can a sleep tracker detect sleep apnoea?
Some devices flag signs of breathing disturbance, but these are screening signals rather than diagnoses and can be wrong in both directions. A persistent flag alongside loud snoring or daytime sleepiness is worth taking to a GP or doctor.
Can sleep tracking make sleep worse?
It can. Clinicians use the term orthosomnia for sleep that deteriorates as someone tries harder to optimise their tracker's numbers, because anxiety about sleep reliably damages it. If the morning score changes how you feel about your day, stop checking it daily.