Sleep Trackers Part 2: How Accurate Are They?

In Part 1 of this series, we looked at the more human side of sleep tracking: sleep scores, sleepmaxxing (trying to perfect your sleep), the pressure to optimise everything, and what can happen when we start trusting an algorithm more than our own experience.

But there's another perfectly reasonable question:

Which sleep tracker is actually the most accurate?

If a watch, ring or sensor is going to tell you how long you slept, how much deep sleep you had, whether you got enough REM (rapid eye movement) sleep and how well your body recovered overnight, it's fair to ask how much we can trust those numbers.

I really wish the answer was as simple as naming one device as ‘the best’. Unfortunately, it isn't.

Different devices are better at different things. Some are reasonably good at working out when you fell asleep and woke up. Some measure heart rate well. Some are better at estimating light sleep, deep sleep and REM sleep. A few also have regulated features designed for a medical purpose: looking for patterns that may suggest sleep apnoea, where breathing is repeatedly interrupted during sleep.

But no consumer tracker measures every part of sleep accurately, and none gives us the same information or accuracy as a proper clinical sleep study.

So perhaps the better question, even if it's overly wordy, is:

Which tracker is most accurate for the thing you actually want to measure, and what are you planning to do with that information?

Sleep tracking is now part of everyday life

Sleep tracking isn't just for elite athletes or dedicated ‘biohackers’ trying to fine-tune their bodies. In a UK poll carried out in 2024, with a sample chosen to reflect the wider population, 22% of adults said they had used an app or device to track their sleep. Another UK poll in 2025 found that 30% of adults owned a smartwatch, 18% used a fitness tracker and 2% owned a smart ring.

That doesn't mean everyone who owns one checks their sleep data every morning. But it does show how quickly wearable health technology has become part of everyday life.

Sleep trackers now come in several forms:

• watches and wristbands, including Apple Watch, Fitbit, Garmin, WHOOP, Samsung Galaxy Watch and Google Pixel Watch;
• smart rings, including Oura, Ultrahuman and Leep;
• under-mattress sensors, including the Withings Sleep Analyzer;
• bedside sensors and phone apps, such as SleepScore Max and Sleep Cycle;
• headbands and other devices that try to measure brain signals more like those used in clinical sleep testing.

These devices aren't all doing the same job. They use different sensors, different algorithms and different ways of turning that data into a score. That makes them hard to compare, especially when the algorithms and sleep scoring tend to be closely guarded commercial secrets.

What are we comparing them with?

The most thorough standard test for measuring sleep is polysomnography, usually shortened to PSG. This is the overnight test used in a sleep laboratory. It records brain activity, eye movements, muscle tone (how relaxed or active your muscles are), airflow, breathing effort, oxygen levels and heart rhythm.

A trained sleep professional reviews the recording in 30-second sections and labels each one as time awake, light sleep, deep sleep or REM sleep. PSG gets much closer to sleep itself because it records the brain signals used to define the different stages. Without those brain signals, it's just guesswork.

Even with those brain signals, interpreting sleep stages isn’t always straightforward. Believe it or not, two trained professionals can look at the same 30-second section and disagree about which stage it shows. So while PSG is our best reference for measuring sleep stages, it still involves human judgement and isn’t a perfect yardstick.

Consumer trackers work very differently from PSG, though. Most watches and rings use some combination of:

• movement;
• pulse rate;
• pulse rate variability (small changes in the time between pulse beats);
• skin temperature;
• blood oxygen estimates;
• breathing-related signals.

Some devices measure only a couple of these things; others measure several. The device puts those signals through its own ‘black box’ algorithm, whose workings we can't see, to estimate whether you were asleep and, often, which stage you were in.

The important distinction is this:

A clinical sleep study measures the brain signals that define sleep stages. A consumer tracker makes an educated guess about those stages using signals from your wrist, finger or mattress.

It may be a very educated guess, but it's still a guess.

What does the latest evidence tell us?

The research is starting to give us a fairly consistent picture. A 2025 meta-analysis (a study that pools the results of several other studies) brought together 24 studies involving 798 people. It compared consumer devices worn on the wrist with polysomnography.

On average, the trackers:

• underestimated total sleep time by around 17 minutes;
• underestimated sleep efficiency (the percentage of your time in bed that you spend asleep) by roughly five percentage points;
• gave different estimates from PSG for how long people were awake during the night.

Those are averages. The errors vary between devices, people and nights. A tracker can overestimate sleep one night and underestimate it the next. Average those nights together and the result can look reasonably good, even if the reading for an individual night was way out.

A 2026 review of consumer sleep devices used in everyday settings reached a similar conclusion. Trackers were fairly accurate for broad measures such as time in bed and total sleep time. But they were less reliable at estimating broken sleep, sleep efficiency and individual sleep stages.

A useful rule of thumb is this:

Trackers are better at estimating roughly when and how long you slept, than they are at telling you exactly what happened during your sleep.

Good at spotting sleep, less good at spotting wakefulness

One of the most consistent findings is that consumer trackers are generally good at recognising when you're asleep. In many studies, when someone was actually asleep, the tracker correctly recorded that time as sleep more than 90% of the time. That sounds impressive, and it is useful.

The trouble is, many trackers are much less accurate at recognising when you're awake. Most watches and rings use a lack of movement as one clue that you've fallen asleep. If you're lying quietly in bed, wide awake but barely moving, your device often thinks you're asleep.

That can be a real problem if you have insomnia. You might spend a long time lying still, trying to drop off, while your tracker records much of that as time spent asleep. Then you open the app in the morning and see a result that bears very little resemblance to the night you remember.

It can happen the other way round too: you move briefly in your sleep, and the device records it as time awake. These errors aren't always predictable. A tracker may overestimate short spells of wakefulness but underestimate long stretches of time lying quietly awake. Accuracy also tends to get worse when sleep is more broken or disturbed.

So the device may be least accurate on the very nights when you most want a clear answer.

Sleep stages are still the least reliable numbers

The graph showing your sleep stages is often the most eye-catching part of a tracker app. It shows a neat journey through:

• light sleep;
• deep sleep;
• REM sleep;
• wakefulness.

It looks very precise. Unfortunately, working out which sleep stage you're in is still one of the least reliable parts of consumer sleep tracking.

A 2024 study compared the Oura Ring Gen3, Fitbit Sense 2 and Apple Watch Series 8 with polysomnography in 35 healthy, young adults. All three devices correctly recognised at least 95% of the time people were asleep.

When it came to telling light sleep, deep sleep and REM sleep apart, though, the results were less impressive. The devices correctly recognised roughly 50% to 86% of the time spent in each stage, depending on the device and the stage being measured.

When the researchers compared the average nightly totals across the group:

• Oura’s average amounts of time awake, light sleep, deep sleep and REM sleep were not statistically significantly different from PSG;
• Fitbit Sense 2 overestimated light sleep by around 18 minutes and underestimated deep sleep by about 15 minutes;
• Apple Watch Series 8 overestimated light sleep by around 45 minutes and underestimated deep sleep by about 43 minutes.

Those are averages, and averages can hide much larger errors in individual readings. If a tracker records too much deep sleep for one person and too little for another, those errors can cancel each other out. The group average may look reassuringly close to PSG even when some individual readings are quite a long way out.

In fact, the study found poor agreement between Oura and PSG for individuals’ total amounts of deep sleep and REM sleep, despite those similar group averages. So “no significant difference” does not mean “just as accurate as a clinical sleep study”.

And even getting the total minutes roughly right doesn’t mean a tracker correctly identified when each stage happened. Two devices worn by the same person can still produce very different sleep-stage graphs. Neither measures the brain directly; each uses its own algorithm to make its best guess.

There is also a funding caveat here: the study received funding from Oura, and its lead author disclosed membership of Oura’s medical advisory board and consulting fees from the company. That doesn’t invalidate the findings, but it is a potential conflict of interest to bear in mind.

Another study from 2022 gives us a useful reality check. Six devices were compared in 53 healthy adults during one night in a sleep laboratory: Apple Watch Series 6, Garmin Forerunner 245, Polar Vantage V, Oura Ring Gen2, WHOOP 3.0 and Somfit. Somfit was a small adhesive patch worn on the forehead. It recorded electrical signals related to brain activity, eye movements and muscle tone, so it was closer to a clinical sleep sensor than the watches and ring. Compared with PSG, Oura agreed with the simpler sleep-versus-wake classification in 89% of 30-second sections. But when the researchers asked whether each section showed wakefulness, light sleep, deep sleep or REM sleep, agreement fell to 61%. Somfit reached 65%, WHOOP 60%, Apple Watch 53%, Garmin 50% and Polar 51%.

This helps put the 2024 Oura study into perspective. Taken together, the studies suggest that Oura may be among the better consumer devices for estimating sleep measures, but they do not show that it measures sleep stages as accurately as PSG. 

Why one night’s score matters less than you think

Sleep naturally varies from night to night. It isn't supposed to look exactly the same every morning.

An analysis that combined data from 2,404 healthy adults, covering more than 26,000 nights, found that the amount of sleep a person got varied quite a lot from one night to another. Depending on how sleep was measured, the typical variation from one night to the next was well over an hour. Sleep efficiency also changed naturally between nights.

That doesn't mean you should ignore every big change. It means that one shorter, longer, lighter or more restless night doesn't mean that something has gone wrong.

Sleep varies in response to all sorts of things, including:

• how much sleep you had on previous nights;
• physical activity;
• mental demands;
• stress;
• illness;
• alcohol;
• medication;
• hormonal changes;
• light exposure;
• temperature;
• travel;
• everyday life.

Your body is constantly adjusting your sleep. Your tracker, meanwhile, may turn that normal variation into a red score, an amber warning or a message saying your recovery is ‘below target’. Suddenly, a normal bodily change can feel like a failure or a danger.

One night tells you something about that particular night. On its own, it says very little about your overall sleep health. Patterns over several weeks or months are much more useful.

Which measurements are most useful?

It helps to put tracker measurements into three broad groups.

Usually more useful over time

• Roughly how long you sleep
• Bedtime and wake time
• How consistent your sleep schedule is
• Time spent in bed
• Changes in your resting heart rate over time
• Broad patterns linked with travel, alcohol, illness, stress or changes in routine

Potentially useful, but worth treating carefully

• Sleep efficiency: the percentage of your time in bed spent asleep
• Time spent awake during the night after first falling asleep
• Sleep latency: how long it takes you to fall asleep
• Heart rate variability (HRV) trends: changes in the variation between heartbeats
• Breathing rate: how many breaths you take per minute
• Readiness and recovery scores: the device's estimate of how recovered and ready for activity you are

Treat these with the most caution

• The exact number of minutes of deep sleep
• The exact number of minutes of REM sleep
• A single night’s sleep-stage graph
• One-off HRV changes
• One-off recovery scores
• An overall sleep score based on the manufacturer's own formula

The more precise a number looks, the easier it is to assume it must be accurate. But a very specific number can still be wrong. And even an accurate measurement isn't necessarily important for your health.

So, what next?

Sleep trackers can help you spot broad patterns in when and how long you sleep. But the finer details, especially those exact deep sleep and REM minutes, should be treated with more caution. And because sleep naturally changes from night to night, looking at patterns over several weeks will usually tell you more than picking apart one disappointing score.

But do some devices deserve more trust than others? In Part 3, we’ll compare the popular watches, rings and under-mattress sensors to see what the research actually supports.

And if you missed Part 1 of this series, head back for a look at our relationship with sleep tracking, and what happens when the pursuit of a better score starts getting in the way of better sleep.

If sleep has become something you dread, monitor or constantly try to control, Retrain Your Sleep can help you rebuild confidence and sleep more naturally using evidence-based CBT-I techniques. 

Click here to learn more about Retrain Your Sleep.

Dr Eidn Mahmoudzadeh is a Sleep Doctor and NHS GP with over 20 years of clinical experience. He is a regular guest on podcasts, a frequent contributor to press and media discussions, and a keynote speaker at many prestigious industry and NHS events.

References

Chee MW, Baumert M, Scott H, et al. World Sleep Society recommendations for the use of wearable consumer health trackers that monitor sleep. Sleep Medicine. 2025;131:106506.

Chinoy ED, Cuellar JA, Huwa KE, et al. Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep. 2021;44(5).

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).

Landvatter J, Ryan JT, Depner CM, Baron K. Real-world use of consumer sleep devices: a rapid review. Chest. 2026;169(4):1101–1110.

Lee YJ, Lee JY, Cho JH, Kang YJ, Choi JH. Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis. Journal of Clinical Sleep Medicine. 2025;21(3):573–582.

Messman BA, Wiley JF, Yap Y, et al. How much does sleep vary from night-to-night? A quantitative summary of intraindividual variability in sleep by age, gender, and racial/ethnic identity across eight-pooled datasets. Journal of Sleep Research. 2022;31(6).

Robbins R, Weaver MD, Sullivan JP, et al. Accuracy of three commercial wearable devices for sleep tracking in healthy adults. Sensors. 2024;24(20):6532.

Miller DJ, Sargent C, Roach GD. A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults. Sensors. 2022;22(16):6317.

YouGov. From sex to exercise, here are the things that Brits are most likely to track using apps and tech. Published 25 March 2024.

YouGov. Nearly half of Britons concerned about wearable tech privacy. Published 17 March 2025.