Your Wearable’s Deep Sleep Number Is a Guess

#

What the devices agree on, and what they don’t

#

Total sleep duration is the reliable part. Across devices, total time asleep typically lands within about an hour of consensus, and for detecting sleep versus wake, validation studies put sensitivity at 95% or above for all major devices. If you want to know whether you slept six hours or eight, your wearable knows.

Stage breakdown is where it falls apart. Wearing an Apple Watch, Oura Ring, Whoop, and Fitbit simultaneously produces two camps on deep sleep — roughly 47 minutes from one pair, 70–81 from the other. REM inverts the grouping: the device most conservative about deep sleep can be the most generous about REM.

That inversion is the tell. These aren’t measurements with differing precision. They’re different proprietary algorithms inferring stages from heart rate, movement, and temperature — inputs that don’t directly observe brain activity, which is what sleep staging actually requires.

What the validation data says

#

Here’s where the popular framing needs correcting. It’s often claimed that Apple’s staging comes closest to laboratory truth, so its lower deep-sleep numbers must be the accurate ones and the higher numbers flattering.

The peer-reviewed comparison against polysomnography — the gold standard, measuring brain activity directly — found the opposite ordering. In a Brigham and Women’s Hospital evaluation of four-stage classification, Oura scored highest (Cohen’s kappa 0.65), Apple Watch next (0.60), Fitbit last (0.55).

On deep sleep specifically, sensitivity ran 79.5% for Oura, 61.7% for Fitbit, and 50.5% for Apple Watch. The devices also err in characteristic directions: Fitbit overestimated light sleep and underestimated deep sleep, while Apple underestimated both wake and deep sleep and overestimated light.

So a low deep-sleep reading isn’t evidence of rigour. It can simply be an underestimate. The honest summary is that stage-level numbers from any wrist or finger device carry error large enough that comparing your figure to a friend’s on a different brand is meaningless.

The failure mode that matters more than accuracy

#

There’s a problem with these devices that has nothing to do with measurement error, and it’s the one worth taking seriously.

Once you’ve learned your patterns, daily data stops informing you and starts overriding you. You wake feeling genuinely good, check a recovery score of 67, and feel worse — not because anything changed in your body, but because a number disagreed with you. The score wins the argument against your own experience.

That inversion is the real cost. The point of tracking is to inform judgment; past a certain point it replaces judgment, and the thing being replaced — how you actually feel — is a signal with far better validation than any consumer algorithm.

Someone who has tracked daily for five years has already extracted nearly all the available insight. Continuing produces diminishing information and increasing interference.

What each metric is actually good for

#

Sleep duration and trends: trust these. They’re well validated and the trend is what matters anyway.

Stage breakdown: read directionally at most. Useful for noticing that your deep sleep has trended down over a month. Useless as a nightly score, and not comparable across brands.

HRV and recovery: act only on extremes. Wrist and finger measurement isn’t precise enough for daily decisions. A chest strap is the better source if you genuinely need this. A single low reading is noise; a week of unusually low readings alongside feeling terrible is worth attending to.

Steps: directionally correct is fine. Hip-worn pedometers are more accurate because they sit near your centre of mass, but for a metric you’re using to nudge behaviour, “roughly right” does the job.

Choosing one, if you want one

#

The practical differences are mundane and matter more than accuracy claims. Battery life — a device needing daily charging is the one most likely to miss nights. Form factor — a bulky wrist unit gets in the way during lifting; a ring gets removed and then isn’t measuring. Subscriptions — some charge monthly, some don’t. And app quality varies enormously, which matters because the app is what actually changes behaviour; a daily journal that surfaces “you drank alcohol and slept badly” does more than a precise number nobody acts on.

None of these devices is a medical instrument. They’re behaviour-change tools with a measurement layer attached, and they’re genuinely good at that job. Awareness creates accountability, and that’s real value — as long as the tool stays advisory.

Summary

#

Consumer wearables reliably measure sleep duration and sleep-versus-wake, with sensitivity at or above 95%, but their sleep-stage breakdowns diverge wildly — a 76% spread on deep sleep between devices on the same night. Validation against polysomnography ranks Oura highest for four-stage classification and Apple lowest for deep-sleep sensitivity at 50.5%, so a conservative deep-sleep number reflects a device that underestimates rather than one that’s more honest. The larger risk isn’t inaccuracy, it’s letting a score override how you actually feel.

What to actually do:

1. Trust duration and trends. Ignore the nightly stage breakdown.
2. Never compare your deep sleep to someone on a different brand — you’re comparing algorithms, not sleep.
3. Act on HRV only at extremes, and use a chest strap if you need real precision.
4. If you wake feeling good, you feel good. Don’t let a recovery score talk you out of it.
5. Consider taking a break if you’ve tracked for years — you’ve already learned your patterns.
6. Pick on battery, comfort and subscription, not on advertised accuracy.

Sources & further reading

#

• Accuracy of three commercial wearable devices for sleep tracking in healthy adults, Sensors (2024) — MDPI / PMC
• Brigham and Women’s Hospital validation of Oura, Apple Watch and Fitbit sleep staging against polysomnography — summary / Sleep Review
• A performance validation of six commercial wrist-worn wearable sleep-tracking devices for sleep stage scoring compared to polysomnography — PMC
• Accuracy of Fitbit Charge 4, Garmin Vivosmart 4, and WHOOP versus polysomnography: systematic review — PMC
• Accuracy of 11 wearable, nearable, and airable consumer sleep trackers: prospective multicenter validation study — PMC