Do sleep trackers get it right?
If you log sleep next to your mood, you have probably wondered how far that number can be trusted. It matters. The research in what the evidence shows on sleep and sex drive rests on carefully measured sleep, not on a watch's guess. So here's what sleep tracker accuracy studies found when devices were checked against a lab recording. The short version: roughly right on totals, weaker on wake time and stages, and uneven from device to device.
What a sleep tracker is compared against
The yardstick is polysomnography, or PSG. It is an overnight lab recording of brain waves, eye movement and more, scored in 30 second chunks by trained staff. A validation study puts a device on someone's wrist or bed while PSG runs, then compares the two.
Two words show up in the results. Bias is the average gap: a device that reads 17 minutes short has a bias of minus 17. Agreement is how well the two move together from person to person, often reported as an ICC, where closer to 1 is better.
What the 2025 meta-analysis found
A 2025 meta-analysis in the Journal of Clinical Sleep Medicine pooled 24 studies and 798 participants who wore consumer wrist devices (Fitbit, Garmin, Apple Watch, WHOOP and others) during PSG nights. On average, devices differed from PSG like this:
| Measure | Average gap vs PSG |
|---|---|
| Total sleep time | 16.9 minutes lower |
| Sleep efficiency | 4.7 percentage points lower |
| Sleep latency (time to fall asleep) | 2.6 minutes higher |
| Wake after sleep onset | 13.3 minutes higher |
Those are averages, and the studies behind them disagreed a lot. The heterogeneity statistic reached 84% for sleep efficiency. Some subgroups looked better: Fitbit's wake time did not differ significantly from PSG, and neither did sleep latency for the other brands. This review did not look at sleep stages. The authors concluded the devices are less reliable than PSG on these measures but can still be useful for tracking general sleep patterns.
Total sleep time vs wake time in a single-night lab study
A 2024 study in Sensors put five devices and a research actigraph through one lab night with healthy adults aged 18 to 30. Total sleep time held up for most. The Fitbit Inspire HR read 9 minutes short, the Versa 2 about 11 short, the Oura Ring Gen 2 about 14 short, and the Withings mat 8 minutes long. Agreement with PSG was good, with ICCs from 0.77 to 0.86.
The Garmin Vivosmart 4 was the exception: 42 minutes long, ICC 0.38.
Wake time was the problem. For time awake after falling asleep, every device landed between 0.08 and 0.32 on the ICC scale, which the authors describe as poor. Even the research actigraph only reached 0.47.
Sleep stages are the weak spot
A 2023 multicenter study in JMIR mHealth and uHealth tested 11 trackers against PSG in 75 people: five wearables, three bedside or under-mattress devices and three phone apps. Four-stage agreement ran from 0.26 to 0.69 on the macro F1 score (1 is perfect), depending on the device.
Other gaps were just as wide. Time-to-fall-asleep bias ranged from under a minute (Apple Watch 8) to about 39 minutes (Google Nest Hub 2). Sleep efficiency bias ran from about minus 3.5 to plus 12.8 percentage points. The authors also noted that wearables drifted more on sleep efficiency and bedside devices on sleep latency.
The 2024 lab study found the same pattern at the stage level: light sleep agreement was modest (ICC 0.19 to 0.42), and the authors report that for deep sleep, most devices overestimated shorter amounts and underestimated longer ones.
What the studies could not tell us
Every study has edges. The 2024 study used one night per person and healthy young adults. The 2023 study ran at two sites in Korea, and the sites differed in how much data they collected and in who took part. Lab nights are not home nights. Devices also update, so a 2023 model is not today's. And the meta-analysis could only pool what past studies measured.
What to do with a number that is only roughly right
The studies point one way: a tracker is better at "about how long" than at "which stage". So read a trend across many nights, not one night's deep sleep figure. If you also log how you feel each morning, you can compare the two over weeks and see your own pattern. It helps to keep it light, as in tracking your body without obsessing over the numbers.
The bottom line
Across three studies, consumer trackers were often in the right area for total sleep time, less dependable for time awake and for stages, and uneven from one device to the next. A rough number is still a number. Just know how rough it is.
Frequently Asked Questions
Are sleep trackers accurate?
For total sleep time they are often close. The 2025 meta-analysis put the average wrist-worn gap at about 17 minutes below polysomnography. For wake time and sleep stages they were less dependable, and results differed a lot by device.
What is polysomnography?
An overnight lab recording of brain waves, eye movement, muscle tone and more, scored in 30 second epochs by trained staff. Validation studies use it as the reference.
Do all trackers perform the same?
No. In the multicenter study, four-stage agreement ranged from 0.26 to 0.69 across 11 devices, and the meta-analysis found different results by brand.
Should I trust the sleep stage breakdown?
The studies suggest holding it loosely. In the lab studies covered here, agreement for deep sleep and wake time was weaker than for total sleep time.
Questions
- Are sleep trackers accurate?
- For total sleep time they are often close. The 2025 meta-analysis put the average wrist-worn gap at about 17 minutes below polysomnography. For wake time and sleep stages they were less dependable, and results differed a lot by device.
- What is polysomnography?
- An overnight lab recording of brain waves, eye movement, muscle tone and more, scored in 30 second epochs by trained staff. Validation studies use it as the reference.
- Do all trackers perform the same?
- No. In the multicenter study, four-stage agreement ranged from 0.26 to 0.69 across 11 devices, and the meta-analysis found different results by brand.
- Should I trust the sleep stage breakdown?
- The studies suggest holding it loosely. In the lab studies covered here, agreement for deep sleep and wake time was weaker than for total sleep time.
Sources
- https://pmc.ncbi.nlm.nih.gov/articles/PMC11874098/
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10654909/
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10820351/
This is general information, not medical advice. The disclaimer says more.


