Experiment guides → When sleep goes wrong → Tracker accuracy
How accurate are sleep trackers?
Quick answer: accurate enough for one thing, unreliable for the things people obsess over. In a lab comparison against polysomnography — the medical gold standard — seven consumer devices were good at recognising when you were asleep, noticeably worse at recognising when you were awake in bed, and only modestly accurate at sleep stages: "deep" and "REM" minutes are estimates, not measurements. A separate case series describes orthosomnia: people whose sleep got worse because they were chasing the tracker's score. The useful stance is the one sleep researchers take — trackers are fine for trends in duration and timing, poor for nightly verdicts, and worthless if the number makes you anxious. The experiment below tests whether yours is helping or ruling you.
"For 14 mornings I will rate my sleep before looking at my tracker."
What the research says
Seven devices vs polysomnography: sleep yes, wake no, stages so-so. Chinoy and colleagues had 34 healthy adults sleep in the lab wearing seven consumer trackers alongside full polysomnography. Devices detected sleep with high sensitivity but detected wake poorly — they overestimated total sleep and underestimated wake after sleep onset — and their sleep-stage classification was only moderately accurate. Some devices were clearly better than others, and all were worse on nights with more waking (Chinoy et al., 2021).
Orthosomnia: when the score becomes the problem. Baron and colleagues coined the term from patients whose pursuit of tracker-perfect sleep was itself causing insomnia — spending longer in bed to raise the score, worrying over "light sleep", disputing their own experience because the device disagreed. The tracker's inaccuracy at detecting wake makes this worse: it can report a fine night after a bad one, and vice versa (Baron et al., 2017).
What researchers recommend: know the limits. A 2024 state-of-the-science statement on wearables in sleep research sets out how the field uses them: validated for sleep-wake timing and duration trends, not for clinical diagnosis or staging, with device-specific performance and a strong caution against treating outputs as ground truth. If the scientists don't trust the deep-sleep number, neither should the rest of us (de Zambotti et al., 2024).
What this evidence doesn't say
The validation study tested specific devices in 2020–21; algorithms update and newer models may do better, though the underlying signal (movement plus heart rate) hasn't changed. Lab nights differ from home nights. And "how rested you feel" is a subjective measure with its own biases — the point isn't that feelings beat sensors, but that neither should overrule the other without a look.
Run the rate-first experiment
- Rate before you read, for 14 mornings. Before touching the phone or watch: sleep quality 1–10, how rested 1–10, and how long you think it took to fall asleep. Write them down. Then look at the tracker.
- Log both numbers side by side. Your rating and the tracker's score. Note the mornings they disagree — a great score after a night you'd call awful, or the reverse.
- Make the pact. "For 14 mornings I will rate my sleep before looking at my tracker." The sequence is the experiment; a score seen first rewrites the memory of the night.
- Watch for the orthosomnia signs. Did a low score make you feel worse than you had? Did you stay in bed longer to improve it? Did you argue with your own experience? Each is a note.
- Decide what the tracker is for. If your ratings and the score mostly agree, the tracker is a reasonable trend tool — keep it, and use your rating as the verdict on any experiment. If they often disagree or the score drives your mood, take it off for the next 14 nights and see whether you sleep better without it.
What to expect
Most people find agreement on duration and rough timing and disagreement on "quality" — the tracker's stage-based score often bears little relation to how the night felt, which is what the validation data predict. A surprising number discover the score has been setting their morning mood. Either way, you finish knowing which number to trust for the experiments on this site: the rating you gave before you looked.
Who should skip this
No risk. If you use a tracker for a medical reason (suspected apnoea screening, a clinician's request), keep using it as directed — this experiment is about the consumer score, not clinical monitoring.
Frequently asked questions
Are sleep trackers accurate for deep sleep?
Not very. Against polysomnography, consumer devices classified sleep stages with only moderate accuracy; "deep" and "REM" are inferred from movement and heart rate, not measured from brain activity. Treat stage minutes as a rough guess, not a diagnosis.
Can a sleep tracker make sleep worse?
Yes — sleep clinicians describe orthosomnia, where chasing a better score leads to more time in bed, more worry and worse insomnia. If the score affects how you feel about the night, that's the sign to rate first, read second, or take it off.
Should I trust my sleep tracker or how I feel?
For duration and timing trends over weeks, the tracker is decent. For whether last night was good, how you feel is at least as valid and doesn't have the wake-detection problem. Researchers use wearables for trends and warn against nightly verdicts; the same rule serves you.
Related experiments
Sources
- Chinoy ED, Cuellar JA, Huwa KE, et al. Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep. 2021;44(5):zsaa291. PubMed 33378539
- Baron KG, Abbott S, Jao N, Manalo N, Mullen R. Orthosomnia: are some patients taking the quantified self too far? J Clin Sleep Med. 2017;13(2):351–354. PubMed 27855740
- de Zambotti M, Goldstein C, Cook J, et al. State of the science and recommendations for using wearable technology in sleep and circadian research. Sleep. 2024;47(4):zsad325. PubMed 38149978
Track this experiment in N of 1
A few baseline nights, then the change — and your own mornings deliver the result. No streaks, no guilt. Your first experiment is free.
Downloadon the App StoreThis page is general information about published research, not medical advice. If you have a sleep disorder, are pregnant, or take medication, talk to your doctor before changing your routine.