Nearly half of adults in some surveys have used a sleep tracker, a smartwatch, a ring, or an app that promises to decode your nights. The pitch is compelling. Measure your sleep, find the patterns, optimize the numbers, wake up better. Self knowledge as self improvement.
But there is a paradox at the heart of sleep tracking. The devices are decent at measuring some things and unreliable at the things people obsess over most. And research suggests that for a meaningful minority, the tracker makes sleep worse. I dug into the accuracy studies and behavioral research to sort out when tracking helps and when it backfires.
What Trackers Actually Measure
To understand the evidence, you need to understand what is inside the device on your wrist. Consumer sleep trackers estimate sleep using actigraphy, movement sensing via accelerometer, and photoplethysmography, heart rate and heart rate variability via optical sensor. From those two signals, algorithms infer when you fell asleep, when you woke up, and how much time you spent in light, deep, and REM sleep.
Clinical sleep measurement, polysomnography, works completely differently. A sleep lab records brain wave activity with EEG, muscle tone with EMG, and eye movements with EOG. Sleep stages are defined by brain activity. A wrist device has no access to your brain. It is guessing at brain states from wrist movement and pulse, which is a fundamentally harder problem than the marketing suggests.
This distinction explains everything that follows. When your tracker reports “14 percent deep sleep,” it is not measuring your deep sleep. It is estimating it from proxy signals, with error bars the app never shows you.
The Accuracy Data: Good at Duration, Bad at Stages
How large are those error bars? A study published in 2025 and indexed on PubMed Central (PMID 38873908) tested two commercial trackers, a Withings device and a Fitbit, against polysomnography in 65 patients referred for sleep studies. The results were sobering.
Both devices estimated sleep start time reasonably well, with median errors of just a few minutes. Both overestimated total sleep time. But for sleep architecture, the actual staging of light, deep, and REM sleep, the agreement with polysomnography was poor. The kappa statistic for epoch by epoch concordance was 0.12 for the Withings device and 0.07 for the Fitbit. For context, a kappa of 1.0 is perfect agreement and 0 is chance. These devices were barely above chance at labeling what stage of sleep you were in. The authors concluded that commercial trackers are not reliable for sleep architecture in patients with sleep complaints and should not replace clinical measurement.
The broader literature agrees. Wearables detect sleep versus wake with roughly 85 to 95 percent agreement and usually land within 15 to 30 minutes on total sleep time. But stage estimates can be off by 30 to 60 minutes nightly. The numbers people fixate on, deep sleep minutes, REM percentage, sleep score, are the least reliable the device produces.
Trackers also overstate sleep efficiency and mistake quiet wakefulness for light sleep. Lie still in bed awake and a movement based device may log you as asleep. Your great night might have included 40 minutes of motionless insomnia.
Do They Change Behavior? The Norwegian Data
Accuracy is only half the question. A device does not need to be perfect to be useful, if using it changes behavior for the better. So does tracking actually improve sleep?
A study from the University of Bergen surveyed more than 1,000 Norwegian adults about their tracker use. Forty six percent had used sleep apps or trackers. Only 15 percent reported that tracking actually improved their sleep. Meanwhile, 18 percent said the devices made them more anxious about their rest. The benefit was concentrated, the anxiety was widespread.
The age split was stark. Among users aged 18 to 35, 23 percent experienced tracker induced stress, compared to just 2.4 percent of users over 66. Younger users engage more obsessively with their data, turning normal nightly variation into a source of anxiety.
Researchers at the University of Oxford went further and tested causality, not just correlation. They manipulated the sleep scores participants saw, giving some people fake poor sleep feedback. Participants who received the fake bad scores reported worse mood and worse cognitive performance the next day, despite having gotten identical actual sleep to the positive feedback group. The mind believed the numbers more than the body’s own experience. That is a remarkable finding. The score did not just fail to reflect reality. It overwrote it.
Orthosomnia: When the Tracker Becomes the Problem
Sleep researchers gave this phenomenon a name in 2017. Baron and colleagues, writing in the Journal of Clinical Sleep Medicine, coined “orthosomnia” to describe patients whose sleep deteriorated because of their fixation on tracker data. The pattern was consistent: people chasing perfect sleep scores developed anxiety about their numbers, the anxiety increased arousal at bedtime, and the arousal made it harder to fall and stay asleep. The pursuit of perfect sleep produced worse sleep.
Orthosomnia is not yet a formal DSM or ICD diagnosis, but the phenomenon is well documented and taken seriously by sleep medicine professionals. The mechanism is textbook insomnia psychology. Worry about sleep is an established insomnia trigger, and a nightly score gives that worry a fresh target every morning.
The cruelest irony: most fixation centers on stage specific metrics, deep sleep percentage, REM duration, precisely the metrics the accuracy research shows are least reliable. People lose sleep over numbers that are often wrong, worrying about a deep sleep deficit that may not exist.
The fundamentals outperform the gadget. Science backed ways to improve sleep, consistent schedule, morning light, caffeine cutoffs, a cool dark bedroom, have decades of research behind them. Melatonin and magnesium have modest but real evidence. And there are strategies to sleep better without medication that address the behavioral core of insomnia.
How to Use a Tracker Without It Backfiring
None of this means trackers are worthless. Used correctly, they are genuinely useful instruments. The key is changing your relationship with the data.
First, track trends, not nights. A single night’s score is noise. A four week trend in total sleep time or bedtime consistency is signal. Check weekly, not each morning.
Second, use the tracker to test behaviors, not grade yourself. Did cutting afternoon caffeine move sleep onset earlier across two weeks? That behavioral experiment role is where self monitoring has real evidence.
Third, prioritize how you feel over the number. The Oxford study proved the score can override your body’s own accurate reporting. If you feel rested and the app says you slept badly, trust your body.
Fourth, know your psychology. If you check the score before deciding how you feel, or a bad number ruins your morning, the tracker costs more than it gives. Take a break.
The Verdict
Do sleep trackers actually help you sleep better? For most people: a little, indirectly, or not at all, and for nearly one in five, they make things worse. The devices track total sleep time decently and spot multi week trends. They are unreliable at sleep stages, the numbers people obsess over. Only 15 percent of Bergen users reported improvement; 18 percent reported more anxiety. The Oxford experiment showed the numbers can override real experience.
My take: a sleep tracker is a compass, not a judge. Notice patterns across weeks, test one change at a time, then put the score in its place. The behaviors that improve sleep have not changed: consistent schedule, morning light, darkness at night, less caffeine, less alcohol, a real wind down. No wrist device changes that list.
