A proposed class action against Oura is turning what might have looked like a consumer wearables dispute into a much bigger question for healthcare, research and the rapidly expanding market for digital biomarkers. Filed in California on August 20, 2026, the lawsuit alleges that Oura overstated the accuracy of its sleep tracking, including claims around sleep-stage accuracy. Oura disputes the allegations and says its science supports the claims it makes. But the case raises an uncomfortable question for the NHS, universities and life sciences companies increasingly using wearables in observational studies: when does an algorithmic estimate become reliable enough to be treated as evidence?

The 95% Number Is Now Under the Microscope

Oura has built much of its reputation around sleep. Its rings collect signals including heart rate, heart-rate variability, temperature and movement, then use algorithms to estimate when someone is asleep and which stage of sleep they are experiencing. The lawsuit argues that consumers may have been given an overly confident impression of what those estimates can deliver. Clinical sleep staging normally relies on polysomnography, including brain activity, eye movement and muscle activity. A ring worn on a finger does not collect those same signals. The complaint focuses on advertising referring to accuracy as high as 95%. Yet accuracy for distinguishing asleep from awake is not the same as accurately separating REM, light and deep sleep. Different studies, populations, device generations and software versions can also produce very different results. That distinction may prove more important than any single headline percentage.

Oura Is Fighting Back

Oura says it stands behind the science, research and accuracy claims supporting its products. The company says its ring estimates sleep stages using physiological signals including heart rate, HRV, movement, breathing patterns and temperature, and points to peer-reviewed studies comparing wearable-derived sleep measurements with clinical sleep testing. That matters because the lawsuit is not a scientific verdict. No court has ruled that Oura’s technology is inaccurate, and a proposed class has not established that the company misled consumers. There is also published evidence showing that wearable-derived measures can provide meaningful information about sleep. The more important question is therefore not whether wearables can estimate sleep. It is whether the limitations, methodology and meaning of those estimates are communicated clearly enough when the numbers are used by consumers, researchers or healthcare systems.

This Is Bigger Than One Ring

For the NHS, the implications extend well beyond Oura. Wearables are increasingly being used in observational studies, decentralised research, remote monitoring and real-world evidence programmes. A study collecting heart rate, movement or temperature is capturing physiological sensor data. A study collecting ‘deep sleep’, ‘REM sleep’ or a proprietary ‘sleep score’ may instead be analysing an algorithm’s interpretation of those signals. Those are not automatically the same thing. Researchers therefore need to be explicit about whether a variable is directly measured, sensor-derived or algorithm-derived, and whether it has been validated for the patient population being studied. This becomes particularly important in cancer, haematology, respiratory disease, neurological conditions and chronic disease, where medication, pain, fatigue, reduced mobility and age can influence the signals on which wearable algorithms depend. Validation in healthy consumers may not be enough for an NHS patient cohort.

What NHS Studies Should Change Now

The answer is not to abandon consumer wearables. It is to use them more intelligently. NHS researchers should specify the device model, firmware and algorithm version used in a protocol, separate directly measured signals from proprietary derived scores, and avoid describing wearable sleep stages as though they were equivalent to polysomnography. Where sleep is an exploratory endpoint, wearable estimates can be extremely valuable because they provide longitudinal data that would be almost impossible to collect through repeated visits to a sleep laboratory. Where sleep staging is a primary clinical endpoint, however, the threshold should be much higher. Studies may increasingly need validation sub-studies comparing wearable outputs against polysomnography, validated actigraphy or another accepted reference standard. That would not weaken wearable research. It would make it more credible.

The Real Fight Is Over Trust

The Oura case exposes a problem that extends across the entire wearable economy. Healthcare is moving rapidly towards continuous patient-generated data, with AI transforming signals from watches, rings, patches and phones into digital biomarkers used by researchers, pharmaceutical companies and healthcare systems. The next question cannot simply be whether a device produces a number. It must be what that number represents, how it was validated, in which population, using which algorithm and whether it is reliable enough for the decision being made. For NHS observational research, that is the central lesson. The future of wearable science is unlikely to involve fewer algorithms. It will require clearer claims, stronger validation and better evidence behind them.