Is Oura Sleep Tracking Accurate? What the Lawsuit's Numbers Actually Measure

Independent and manufacturer-associated validation evidence for Oura Ring sleep tracking against polysomnography, read against the accuracy figures at the centre of the proposed class action. It does not assess the legal merits of the case.

Both numbers in the Oura lawsuit are real, and they measure different things. The 53.18% figure is four-stage accuracy, meaning how often the ring correctly labelled each 30-second stretch of a night as wake, light, deep or REM sleep, measured in a university sleep laboratory population with diagnosed sleep disorders and other medical conditions (Scientific Reports, 2025). The figure near 95% is sleep detection, meaning how reliably the ring recognises that a person is asleep at all, measured in 96 generally healthy adults, where sensitivity reached 94.4 to 94.5% (Sleep Medicine, 2024). Detecting sleep is a much easier task than dividing it into stages, and healthy sleepers are much easier to score than sleep-clinic patients. The weak spot both studies agree on is wake: in the healthy-adult validation the ring correctly identified wake only 73.0 to 74.6% of the time.

What the lawsuit says

A proposed class action filed in August 2026 accuses Oura of misleading consumers about the accuracy of its sleep tracking, contrasting an advertised accuracy figure of around 95% with a published study that found 53.18% sleep-stage accuracy. Oura has said it stands behind its science and accuracy claims and intends to defend the case. The suit is a proposed class action and has not been decided.

This page does not assess the legal merits. It answers the question the case raises for anyone wearing the ring: what does the published validation evidence show, and where does each number come from.

Where the 53.18% comes from

The 53.18% figure comes from a study that tested three commercial ring trackers, including the Oura, against polysomnography in a university sleep laboratory. The participants were not healthy volunteers. They were a sleep-lab patient population with a diverse set of sleep-related disorders and sleep-unrelated medical conditions, and the study was designed specifically because most earlier validation work had been done on healthy people.

Across all four stages, the Oura ring reached 53.18% accuracy, close to the SleepOn ring at 50.48% and well ahead of the Circul ring at 35.06%. On the simpler task of telling sleep from wake, the Oura reached around 85%, which the authors note is similar to what earlier studies reported in healthy individuals. At the group level the ring was close to the laboratory: average total sleep time differed by less than 12 minutes. The problem the authors emphasise is the individual night. Averages agreed while individual nights were often far off, and the bias was complex enough that it could not be simply corrected (Scientific Reports, 2025).

So the 53.18% is a real, peer-reviewed result. It is the four-stage figure, in a clinical population, and it is the hardest version of the test a sleep tracker can be given.

Where the figure near 95% comes from

The high figures come from a multi-night validation of the Oura Ring Gen3 running Oura's second-generation sleep staging algorithm, in 96 generally healthy men and women aged 20 to 70, covering 421,045 thirty-second epochs scored against ambulatory polysomnography by American Academy of Sleep Medicine criteria.

In that study the ring's sensitivity for sleep was 94.4 to 94.5%, its predictive value for sleep was 95.9 to 96.1%, and its overall sleep-wake accuracy was 91.7 to 91.8%. It did not differ significantly from polysomnography on time in bed, total sleep time, sleep onset latency, wake after sleep onset, or time in light and deep sleep, and it underestimated REM by 4.1 to 5.6 minutes. Per-stage accuracy ranged from 75.5% for light sleep to 90.6% for REM (Sleep Medicine, 2024).

Two details matter when reading those numbers. They describe healthy sleepers, not patients. And per-stage accuracy is calculated one stage at a time, which tends to produce higher figures than a single four-way accuracy across all stages at once, because every correctly rejected epoch counts in its favour. The 90.6% for REM and the 53.18% four-stage figure are not two attempts at the same measurement.

The number neither side leads with: wake

Both studies point at the same weakness, and it is not the one in the headlines. In the healthy-adult validation, the ring's specificity, its ability to correctly identify wakefulness, was only 73.0 to 74.6%, and its predictive value for wake was 66.6 to 67.0%. In practice that means when the ring says you were awake, it is right about two times in three, and some time spent lying awake is likely being counted as sleep (Sleep Medicine, 2024).

For a person with insomnia, or anyone who wakes repeatedly through the night, that is the figure that matters most, and it is the reason a ring can report a good night's sleep that did not feel like one.

What an independent comparison found

A separate study put the Oura Ring Gen3, the Fitbit Sense 2 and the Apple Watch Series 8 under polysomnography on the same nights in 35 adults aged 20 to 50 without a sleep disorder. The Oura ring was the only one of the three that did not differ significantly from the laboratory standard on wake, light, deep or REM sleep, with stage sensitivity of 76.0 to 79.5%. The Apple Watch underestimated deep sleep by 43 minutes, and the Fitbit underestimated deep sleep by 15 minutes (Sensors, 2024).

Another five-device study in 53 young adults found the Oura underestimated light sleep across nights of any length, while REM bias was low for every device tested (Sensors, 2024). Results depend on the sample, the night and the algorithm version, which is part of why single headline figures travel badly.

What this means if you wear one

On the published evidence, the Oura is among the better-performing consumer sleep trackers at telling whether you slept and roughly for how long, particularly if you sleep normally. It is considerably weaker at saying exactly how that sleep divided into stages, it is weakest at recognising time spent awake, and its accuracy falls in people with sleep disorders, which is the group most likely to care.

Averaged over many nights the numbers are more trustworthy than any single night. The clinical study found group averages close to the laboratory while individual nights varied widely, so a trend across weeks is a better use of the data than a verdict on last night (Scientific Reports, 2025).

Common questions