In one sentence
The minutes your device thinks you spent in deep (slow-wave) sleep and in REM sleep, drawn on the night's chart because people expect to see them, and scored nowhere in VT because no device measures them well and no trial has shown a target for them.
Why it is shown and not scored
Sleep has an architecture: deep sleep comes early in the night, driven by the pressure that builds while you are awake, and REM comes late, gated by the body clock toward morning, which is why cutting a night short trims REM more than deep. That much is well established. What is not established is any target: in laboratory recordings of 5,273 healthy adults, deep sleep averaged 20.4 percent of the night and REM 19.0 percent, and neither changed with age (Boulos 2019), which overturned an older meta-analysis that found both falling with age in samples that had not been screened for sleep disorders (Ohayon 2004). The "20 to 25 percent deep sleep" figure that fills sleep apps is a population mean turned into a goal; no trial has assigned anyone a stage percentage and measured an outcome. That is why sleep stages carry zero weight in the Sleep Health Index.
What moves it
Alcohol, strongly: REM is delayed at every dose and reduced across the night at moderate and high doses, while deep sleep increases in the first half (Ebrahim 2013), so a glass of wine can make a tracker's "deep sleep" look better on a worse night. Also: a short night (less REM), a hard training day, a hot room, some medicines, and the device's own algorithm, which is the largest mover of all.
How to measure it well
You cannot, at home. Staging needs the brain's electrical activity; a ring or wrist infers it from movement, heart rate and HRV. In the one ring study against polysomnography, the device agreed with the laboratory on 65 percent of light-sleep epochs, 51 percent of deep and 61 percent of REM, under-reading deep sleep by about 20 minutes and over-reading REM by about 17 (de Zambotti 2019); across the major brands, stage classification runs at roughly 50 to 65 percent epoch accuracy with per-stage errors of 30 to 60 minutes, and the biology adds its own noise, with the deep-sleep share varying night to night by about a fifth of its mean even in laboratory-grade home recordings. The same device on the same person is consistent with itself, so a trend on one device over weeks has some meaning; a night has little.
How Vitality Tracker uses it
Drawn on the night's chart, labeled as a low-confidence estimate, with the device named. Deep sleep and REM are secondary readings in Symptom Radar (weight 0.7 each, lower is the direction it watches), where a fall in both against your own usual on the same device is treated as one sign among several and never as a result on its own. They do not enter the Sleep Health Index or the nightly sleep score.
How far to trust the number
Estimate, and the least trustworthy figure on the sleep screen. Trust the direction on one device over weeks, not the minutes.
Sources
- Boulos MI, et al. Normal polysomnography parameters in healthy adults. Lancet Respir Med 2019;7:533–43. doi:10.1016/S2213-2600(19)30057-830057-8>)
- Ohayon MM, Carskadon MA, Guilleminault C, Vitiello MV. Meta-analysis of quantitative sleep parameters from childhood to old age in healthy individuals. Sleep 2004;27:1255–73. doi:10.1093/sleep/27.7.1255
- Ebrahim IO, et al. Alcohol and sleep I. Alcohol Clin Exp Res 2013;37:539–49. doi:10.1111/acer.12006
- de Zambotti M, et al. The Sleep of the Ring. Behav Sleep Med 2019;17:124–36. doi:10.1080/15402002.2017.1300587
- Vitality Tracker Sleep Scoring Methodology v2.0.1 (§4.1, stage accuracy and night-to-night variability). doi:10.5281/zenodo.23223863.