You open your sleep app before you are fully awake. Deep sleep: 12%. Right underneath, a small grey bar marks the "normal" zone, starting at 20%. And just like that, your day begins with the vague sense that you failed at something while you were unconscious.
You have probably met that 20% figure before. It shows up in forums, in news articles, in the tooltips of tracking apps. It looks solid. It looks medical. But if you go looking for the official recommendation it supposedly comes from, you will not find one. There isn't any.
That does not make the number wrong. It makes it misused: a descriptive average that got promoted to a passing grade.
Where the number actually comes from
The reference behind these percentages is a meta-analysis published in 2004 in the journal Sleep by Maurice Ohayon and colleagues. The authors pooled 65 studies covering 3,577 healthy people aged 5 to 102, all recorded in a lab using polysomnography. Their stated goal was to establish normative sleep values across the human lifespan.
For adults, the result lands somewhere around 18% to 22% of total sleep time spent in slow-wave sleep, as a 2024 review in Frontiers in Sleep summarizes. The famous 20% sits in the middle of that range.
Notice the verb the researchers use: they describe. They measure what healthy sleepers actually do. At no point do they prescribe a target. That distinction matters enormously, and it is exactly the one that got erased on the way to the consumer app.
The Sleep Foundation, a leading public-facing authority on the topic, actually gives a different range: 10% to 20% of total sleep, or roughly 40 to 110 minutes across a seven-to-nine-hour night. And it adds a sentence the apps never quote: there is no consensus on a single, concrete number for how much deep sleep is normal or ideal.
Two serious sources, two ranges that only partly overlap. In one, 20% is the average. In the other, it is the ceiling. A threshold that changes meaning depending on who cites it is not a clinical standard.
How an average became a grade
How does a descriptive statistic turn into a goal? Through a very ordinary shortcut, one you find anywhere humans get measured.
An average describes a group. It says nothing about what is desirable for one person. The average height of an adult is not the correct height. Yet the moment you display an average next to a personal measurement, with a progress bar and a color, the brain reads a grade. Green means you passed. Amber means try harder.
The trouble is that individual variation around these averages is enormous. Two people of the same age, both perfectly healthy, can show very different proportions of deep sleep without either one sleeping better. The 18-22% band is a central tendency, not a lane everyone should stay inside.
The second problem: your watch is not measuring that stage
Suppose for a moment the 20% target did mean something. You would still need to measure it at home. But slow-wave sleep is defined by electrical brain activity, the slow waves visible on an electroencephalogram. Your watch sees neither your brain nor your brainwaves. It observes your movement and the variation in your pulse, then infers a stage through an algorithm.
The quality of that inference has been measured. A 2025 validation published in Sleep Advances by Schyvens and colleagues compared six consumer wrist wearables against polysomnography in 62 adults. Cohen's kappa, which measures agreement beyond chance, ranged from 0.21 to 0.53 across devices: described as fair to moderate agreement.
On deep sleep specifically, the share of epochs correctly classified ran from 47.5% for the weakest of the six to 69.6% for the best. Put plainly: on the very metric that worries you in the morning, the best device in the panel still misplaces nearly one epoch in three, and the weakest more than one in two.
This does not make wearables useless. They distinguish sleep from wake reasonably well, and they track trends across weeks perfectly adequately. But a deep-sleep percentage shown to one decimal place for a single night implies a precision the measurement does not have.
And the number shifts with age — probably
A third layer of nuance. The proportion of deep sleep declines through adulthood, at roughly 2% per decade according to the Ohayon meta-analysis. A single 20% threshold therefore cannot mean the same thing at 25 and at 55.
Except that even this decline is contested. The Frontiers in Sleep review notes that a more recent meta-analysis, restricted to studies using the 2007 AASM scoring manual, does find an age-related drop in total sleep time but no significant reduction in slow-wave sleep. Work from the SIESTA cohort, meanwhile, reports a decrease of about 1.7% per decade in men and no change at all in women.
When specialists disagree about the size of the decline, about whether it survives a change in scoring rules, and about whether it applies to both sexes, it is hard to defend a fixed threshold printed on a phone screen.
When the score keeps you awake
This gap has a measurable cost. Research presented in May 2026 in Sleep by Ronan Goel, Rebecca Robbins, Charles Czeisler and colleagues surveyed 1,280 American adults about their sleep-tracking habits. Almost a third, 32.4%, reported tracking their sleep. Among those, 30.9% screened positive for orthosomnia risk, roughly double previous estimates.
Orthosomnia describes an excessive preoccupation with achieving the "perfect" night as scored by a device, to the point that chasing the score degrades the sleep itself. The loop is simple: the number worries you, the worry tenses bedtime, the tense bedtime damages the next night, and the number gets worse.
There is a particular irony in the least reliable metric on your watch being the one that generates the most anxiety.
What to watch instead
Dropping the threshold does not mean dropping the measurement. It means asking it a different question.
- Your trend, not your night. A single value is swamped by the device's margin of error. An average across three or four weeks, compared against your own earlier weeks, carries real information. Your baseline is you.
- Schedule regularity. Bedtime and wake time come from simple sensors, so they are far more reliable than any stage classification. They are also the levers you can actually pull.
- Total sleep duration. It mechanically governs how much deep sleep you get: a stable percentage of a shorter night yields fewer slow-wave minutes. Lengthening the night stays the most direct lever.
- How you feel on waking. This is the Sleep Foundation's explicit recommendation in the absence of a target number: how rested you feel remains the most accessible indicator.
- What changed the day before. A glass of wine, a late workout, an overheated bedroom. That is where measurement earns its keep, by linking your nights to your days. We unpacked one such mechanism in our piece on alcohol and recovery as seen by wearables.
That is also the spirit of a wellbeing-oriented approach to quantified self: watching trends that belong to you rather than chasing a grade. The trackers and dashboards in Kantise are built to cross your sleep data with what you actually lived through, without stamping a failure color on your morning.
And if the night was short anyway, napping and its effect on cognitive performance remains a documented lever, whatever your wrist reports.
The takeaway
The 20% threshold is not a medical recommendation. It is the midpoint of a range observed in a sleep lab among healthy sleepers, published to describe a population rather than to grade a person. No professional body has set an amount of deep sleep you should reach.
Add the fact that your watch estimates that stage with fair-to-moderate agreement at best, and the morning verdict loses most of its authority. Your sleep did not deteriorate because an algorithm filed a few minutes under the wrong heading.
Ready to observe your nights without grading them? Watch your trends, not your thresholds.
FAQ
How much deep sleep do you actually need per night?
There is no official recommended amount. Lab measurements place adults between 18% and 22% of total sleep time in the Ohayon meta-analysis, while the Sleep Foundation cites 10% to 20%, or roughly 40 to 110 minutes. These are population observations, not individual targets.
My app says 12%. Should I be worried?
A single value tells you very little. Stage classification by a watch shows only fair to moderate agreement with polysomnography, so 12% could reflect measurement error just as easily as a normal variation. If daytime fatigue persists for weeks, that symptom is what warrants medical advice, not the percentage.
Do smartwatches really measure deep sleep?
They estimate it. The stage is defined by brain activity, which wrist devices do not record: they infer it from movement and heart rate. In the Schyvens validation of six devices, the share of deep-sleep epochs correctly classified ranged from 47.5% to 69.6%.
Does deep sleep decrease with age?
Probably, though the magnitude is debated. Ohayon reports roughly 2% per decade in adults, with no significant change after 60. A meta-analysis limited to 2007 AASM scoring finds no significant reduction, and the SIESTA cohort observes a decline in men only.
What is orthosomnia?
It is an excessive preoccupation with achieving the "perfect" night as scored by a device, to the point of harming sleep. In the survey presented in 2026 by Goel and colleagues among 1,280 adults, 30.9% of those who tracked their sleep screened positive for orthosomnia risk.
How can I increase my share of deep sleep?
The more useful question is total sleep: at a stable proportion, a longer night delivers more slow-wave minutes. Consistent schedules and a cool, dark bedroom are the best-established levers. Aiming at a displayed percentage is not one of them.
Sources
- Ohayon MM, Carskadon MA, Guilleminault C, Vitiello MV. Meta-Analysis of Quantitative Sleep Parameters From Childhood to Old Age in Healthy Individuals. Sleep, 27(7), 1255-1273, 2004.
- Ishii T, Taweesedt PT, Chick CF, O'Hara R, Kawai M. From macro to micro: slow-wave sleep and its pivotal health implications. Frontiers in Sleep, 2024.
- Schyvens AM et al. A performance validation of six commercial wrist-worn wearable sleep-tracking devices. Sleep Advances, 6(2), zpaf021, 2025.
- Goel R, Quan S, Weaver MD, Czeisler C, Robbins R et al. Exploring the Prevalence of Sleep Tracking and Orthosomnia in a National Survey of Adults in the US. Sleep, 49(Suppl. 1), A246, 2026.
- Sleep Foundation. How Much Deep Sleep Do You Need?
- Li J, Vitiello MV, Gooneratne N. Sleep in Normal Aging. Sleep Medicine Clinics, 13(1), 1-11, 2018.
Kantise is an observation tool, not a medical device. The information in this article is provided for general information only and does not replace advice from a healthcare professional. If you experience persistent daytime sleepiness, snoring with breathing pauses, or lasting insomnia, please consult a doctor.