The Accuracy Question
Smartwatches claim to track sleep stages (light, deep, REM), but how do they compare to clinical polysomnography? We spent 3 nights in a sleep lab while wearing 6 smartwatches simultaneously, comparing their readings against the gold-standard EEG-based sleep staging.
Results
Apple Watch Series 11: 82% overall sleep stage accuracy. Best at detecting wake periods (94% accuracy) and REM sleep (79% accuracy). Deep sleep was consistently overestimated by 15-20 minutes per night.
Samsung Galaxy Watch 7: 78% overall accuracy. Strong at detecting total sleep time (within 8 minutes of lab measurement) but weaker at distinguishing light from deep sleep stages.
Garmin Venu 4: 75% overall accuracy. Excellent at sleep onset detection (within 3 minutes) and wake detection (91%). The Body Battery feature correlates well with subjective energy levels the following day.
Fitbit Sense 3: 73% overall accuracy. The sleep score algorithm provides useful trend tracking even when individual night staging isn't perfectly accurate.
What's Useful vs What's Not
Useful: total sleep time trends, sleep consistency scores, wake detection, and long-term pattern tracking. Not useful: exact deep/REM minute counts, sleep quality scores based on single nights, and comparisons between different brands (they use different algorithms).
Recommendation
Use smartwatch sleep tracking for trend analysis over weeks and months, not for diagnosing sleep disorders. If you suspect a sleep condition (insomnia, sleep apnea), consult a sleep specialist rather than relying on consumer devices.