What sleep stages are, and how they are properly measured
Sleep is conventionally divided into REM and three non-REM stages, the deepest of which is slow-wave sleep. The reference method for identifying them is polysomnography, which records brain activity by EEG alongside eye movement and muscle tone.
Stage classification is fundamentally a brain-activity measurement. Nothing worn on the wrist or finger records brain activity, which is the central constraint on everything a consumer device reports about your sleep stages.
What a consumer tracker actually records
Wearables infer sleep from movement, heart rate, heart-rate variability, and sometimes temperature or blood-oxygen estimates. An algorithm maps those signals onto stage labels.
Validation studies comparing consumer devices to polysomnography have generally found that they are reasonably good at distinguishing sleep from wake and total sleep time, and considerably weaker at classifying individual stages — with deep sleep among the least accurately estimated.
So a low deep-sleep figure may reflect your sleep, or the limits of inferring brain states from a wrist. Without EEG the device cannot distinguish those two possibilities, and neither can you.
Why 'enough hours' and 'feeling rested' come apart
Sleep continuity matters alongside duration. Fragmented sleep with brief arousals can total eight hours while producing daytime sleepiness, and many of those arousals are too short to be remembered.
Timing matters too. The same eight hours taken at a time misaligned with your circadian phase is not equivalent to eight hours taken in alignment, which is much of why shift work is studied as a distinct exposure.
There are also medical causes of unrefreshing sleep — sleep-disordered breathing being the most common — that a consumer device is not designed to detect and cannot rule out.
Where the evidence is thin
The idea that a specific percentage of deep sleep is optimal is not well supported. Slow-wave sleep varies substantially with age and between individuals, and there is no established target that applies across people.
Research on interventions that increase measured deep sleep is also complicated by the measurement problem: if the estimate itself is the least reliable part of the output, an intervention that appears to move it may be moving the estimate rather than the sleep.
What this research does not establish
- Consumer trackers do not measure brain activity and are weakest precisely at classifying deep sleep — the stage most people ask about.
- No established optimal percentage of deep sleep exists that applies across individuals.
- Persistently unrefreshing sleep can indicate a medical condition such as sleep-disordered breathing and warrants clinical assessment, which no tracker replaces.
Common questions
- Why do I wake up unrefreshed even after a full night?
- Duration is only one component. Sleep continuity, circadian timing, and undetected arousals all affect how restorative sleep is, and brief arousals are often not remembered. Medical causes such as sleep-disordered breathing also produce unrefreshing sleep despite adequate time in bed, and consumer devices are not designed to detect them.
- My sleep score is good but I still feel tired. Which should I trust?
- How you feel is the more direct measure. A sleep score is an algorithmic estimate built from movement and cardiovascular signals, not a measurement of sleep quality, and validation studies show consumer devices are markedly less accurate at stage classification than at total sleep time.
- Why can't I reach deep sleep?
- First consider whether the figure is accurate. Deep sleep is among the least reliably estimated outputs of consumer wearables, because identifying it properly requires EEG. Slow-wave sleep also declines substantially with age and varies widely between people, and no established optimal percentage applies across individuals.
- Are sleep trackers accurate?
- Reasonably good at distinguishing sleep from wake and estimating total sleep time; considerably weaker at classifying individual stages when compared against polysomnography. They are best treated as trend instruments rather than as measurements, and they are not diagnostic devices.