Sleep Tracking Against What Actually Happened
Sleep tracking reports deep sleep, REM and a score out of a hundred. We compared three devices against an observed record of the same nights.
Tech Daily Review Desk
What actually changed
Sleep tracking became the flagship feature of wearables, and the numbers got more specific — stages to the minute, a single score, advice based on both. The confidence of the presentation has run a long way ahead of what a sensor on a wrist can actually know.
A wearable infers sleep from movement, pulse, pulse variability and skin temperature. Those signals separate "asleep" from "awake" quite well and separate the stages of sleep quite badly, because the stages are defined by brain activity that a wrist cannot see. The stage numbers are therefore a model's best guess presented to the minute, and the difference between those two things is the whole subject of this piece.
Why this is worth your attention
Because people change their behaviour on the strength of these numbers, and a subset become anxious about them. A score that is wrong in a reassuring direction is harmless; one that is wrong in an alarming direction costs somebody a night.
How we measured it
Three devices worn simultaneously — two watches and a ring — across 45 nights and nine people, against a reference record assembled from an observed log: timestamped wake events, lights out, and a measured time of first sleep. We did not attempt to reference the stages against a clinical measurement, because we cannot; the stage rows below therefore compare the three devices to each other, which turns out to be the more damning comparison.
What we recorded
| Measure | Watch A | Watch B | Ring | Observed |
|---|---|---|---|---|
| Total time asleep | 7 h 12 m | 7 h 28 m | 7 h 04 m | 7 h 09 m |
| Time to fall asleep | 9 m | 4 m | 14 m | 16 m |
| Wakes over 5 minutes | 1.2 | 0.6 | 2.1 | 2.4 |
| Deep sleep | 1 h 44 m | 58 m | 1 h 21 m | not measurable here |
| REM | 1 h 06 m | 1 h 51 m | 1 h 28 m | not measurable here |
| Nightly score | 74 | 86 | 69 | — |
| Agreement between devices, stages | — | — | — | 41% |
What the data shows
Total sleep time is the one figure to take seriously — all three were within twenty minutes of the observed record, which is a good result for a wrist.
All three consistently missed short wakes and underestimated how long people took to fall asleep, which is precisely the direction that flatters the score.
The stage figures disagreed with each other 59% of the time on the same night on the same body, and watch A reported nearly twice watch B's deep sleep on identical nights.
The same night produced a 74 and an 86 on two devices, which tells you what a score out of a hundred is worth as a number rather than as a nudge.
Caveats
Three things worth holding on to when you meet these figures somewhere else:
- Stage figures are a model's estimate presented to the minute, and the precision of the presentation is not evidence of accuracy
- The score algorithms change with firmware, so a drop in your average may be a software update rather than a change in your sleep
- If a score is making you anxious about sleeping, it is producing the opposite of the thing it measures — turn the notification off
Our advice
Use the trend in total sleep time and ignore everything else. That figure is accurate enough to be useful, it responds to the things you can actually change, and it is the only number in the interface our testing supports. A fortnight of averages tells you something real; a single night's score does not.
Do not act on the stage breakdown at all. Two devices on the same wrist on the same night disagreed about deep sleep by a factor of nearly two, which means at least one of them was substantially wrong every night, and there is no way from the outside to know which. If you have a genuine concern about your sleep, this is a conversation with a doctor rather than a graph.
Where we land
Trust the total, ignore the stages, and treat the score as a mood ring. All three devices tracked time asleep to within twenty minutes and disagreed with each other about sleep stages on three nights in five.
Behind this piece
Every product covered here was bought at retail and lived with for at least a fortnight before a word was written. Battery figures come from a scripted day rather than a manufacturer estimate, and sustained performance is measured warm — after half an hour of load — because that is the state a device spends most of its life in.
Where a firmware update changes a result, the piece is re-tested and the change noted rather than quietly corrected. No review unit arrives with conditions attached, and no score is agreed in advance with anybody.
Where to go next
If you are weighing this against the rest of the wearables we have covered, the category archive lists every review in date order. The closest comparisons we have published are Step Counts Against a Manual Count and The Best Wearables For Battery Life in 2026.
For the reference material behind these figures, the specification pages carry the measured numbers, and the common questions collect what people ask once something is in daily use.
