Sourced table

    Sleep Tracker Validation Scoreboard

    Published sleep/wake and four-stage figures for contactless devices, with the study size, the reference, a link, and the caveat. Wellness reading only.

    Last updated October 31, 2026. Every row has a URL. If we could not open a source, the number is not here.

    DeviceStudy sizeReferenceSleep/wakeFour-stageSourceCaveats
    Sleepal AI Lamp1,022 adultsPSG (lab)92.7%77.2%Physiological Measurement, doi:10.1088/1361-6579/aeb104Peer-reviewed paper cited in our 2026-10-31 teardown. A vendor preprint of the same work (arXiv 2604.16442) reports 92.8% sleep/wake and 78.5% four-stage in the Normal group, with 77.2% four-stage overall. Vendor-side authors. Not an independent trial.
    Withings Sleep117 healthy adultsPSG (home)87%63%JMIR Human Factors 2026Healthy adults at home. Kappa 0.49. Withings also published the weak spots: overestimates light sleep by 1h21m, underestimates deep sleep by 46m.
    Somnofy71 nightsPSG (lab)97% sleep sensitivity, 72% wake specificityNot published in this paperToft et al., PubMed 32853919Healthy, mostly young adults. Authors said more research is needed in older people and in people with sleep disorders. Sold mainly into research and care settings.
    Somnofy (older adults)35 older adults (mean age 70.8)PSG (sleep lab)MCC 0.63 (sleep vs wake)Not published in this paperJMIR 2023, doi:10.2196/46338Matthews correlation, not raw accuracy. Same paper: under-mattress Withings MCC 0.41, Emfit MCC 0.35. All three overestimated total sleep time by more than 90 minutes.
    Komori CareNone publishedNo published validation yetNo published validation yetNo published validation yetKomori Care (pre-launch, 2027)Pre-launch, coming 2027. General wellness only. Not a medical device. This row is here so the table does not imply a hidden score.

    How to read these numbers

    A single accuracy percentage is easy to print and easy to game. Most of a night is sleep, so a device that always says "asleep" still looks respectable. Ask for specificity on wake, Cohen's kappa, who was in the sample, and what the device was compared against. That is the argument in what "95% accurate" actually means.

    Wrist actigraphy on 77 adults scored 86.3% accuracy and 32.9% specificity versus PSG (Sleep, 2013). Seven consumer trackers in 34 healthy young adults hit sensitivity 0.93 to 0.99 and specificity 0.18 to 0.54 (Sleep, 2021). Human scorers agree with each other 82.6% of the time overall, and 63.0% on the lightest stage (JCSM, 2013).

    Lab PSG is not home PSG. A healthy young cohort is not a mixed clinical one. Sleep sensitivity is not four-stage accuracy. If a cell says "not published," we did not fill it in.

    Nest Hub has a WIRED review caveat (overestimated REM, missed wake) and no published PSG percentage in our sources, so it is not a row. Sleepal homepage tiles (99% HR, 91% position) are the vendor's own claims, not a paper, so they are not a row. Sleepiz One+ (SLEEP 2026 abstract 0832) and Albus Home (AJRCCM abstract A30-11) appear in our compared guide but have no public URL in our sources, so they are not a row.

    Cite this page

    Stable URL. Complete-sentence findings, then the HTML if you want a link.

    • Sleepal's peer-reviewed PSG comparison (1,022 adults) reported 92.7% sleep/wake and 77.2% four-stage accuracy (doi:10.1088/1361-6579/aeb104).
    • Withings Sleep's JMIR home study (117 healthy adults) reported 87% sleep/wake and 63% four-stage accuracy (https://humanfactors.jmir.org/2026/1/e77033).
    • Somnofy's 71-night PSG study reported 97% sleep sensitivity and 72% wake specificity (https://pubmed.ncbi.nlm.nih.gov/32853919/).
    • Komori Care has no published validation yet (pre-launch, 2027).
    <a href="https://komoricare.com/sleep-tracker-validation">Sleep Tracker Validation Scoreboard</a>
    https://komoricare.com/sleep-tracker-validation

    Komori Care (coming 2027)

    Komori Care is coming in 2027. It is designed as a contactless bedside sleep-position monitor. No screen or light at the bedside. Designed to keep tracking without a phone in the room. General wellness only. Not a medical device. No published validation yet.

    Komori is coming in 2027. Get the launch note.

    Free waitlist. No validation paper to put in the table yet. We will say so when one exists.

    FAQ

    Why don't these percentages match across devices?
    They are not the same test. Sleepal's journal paper is 1,022 adults versus lab PSG. Withings is 117 healthy adults versus home PSG. Somnofy's 71-night paper reports sensitivity and specificity, not one accuracy number. Read the study size, the reference, and the caveats before you rank a shelf.
    What is four-stage accuracy?
    Agreement with a PSG scorer on wake, light, deep, and REM. Human scorers themselves agree only 82.6% of the time overall (JCSM 2013). A headline that beats that without saying who was tested is a red flag. More in our accuracy-claims post.
    Why is Komori listed with no number?
    Komori Care has no published validation yet. It is pre-launch (2027). We will not invent a percentage.
    Can I cite this table?
    Yes. Stable URL: https://komoricare.com/sleep-tracker-validation. Copy the HTML snippet on the page. Quote a complete-sentence finding and keep the source link.

    Keep reading