Sleep architecture: what your wearable doesn’t tell you
Your Oura ring shows you a deep-sleep number. The labs that established that number used 32-channel polysomnography. Here is what gets lost in translation — and what to actually do about it.
Every consumer wearable that gives you a “deep sleep” number is making a statistical guess. The underlying ground truth — the studies that defined what deep sleep is — comes from in-lab polysomnography (PSG): EEG, EOG, EMG, ECG, oxygen saturation, and airflow, scored in 30-second epochs by a credentialed technician.
Your Oura, your Whoop, your Apple Watch — they are inferring sleep stages from heart-rate variability, movement, and skin temperature. They are good. They are not the same.
What wearables actually measure well
The strongest signal in any wrist or ring wearable is total sleep time. If you trust nothing else from your device, trust the total. The autonomic signature of sleep — bradycardia, smoothed HRV, reduced movement — is unmistakable across all devices we have audited.
Onset latency (how long it takes you to fall asleep) is also generally reliable. So is wake-after-sleep-onset (WASO) for awakenings longer than a few minutes.
Where wearables drift from polysomnography
The further you move from gross physiology into stage-level architecture, the looser the inference becomes. A 2023 validation study comparing Oura Gen 3 to PSG in 96 adults found stage-level concordance ranged from 62% for REM to 79% for wake, with deep sleep landing at 71%. Whoop 4.0 results are broadly similar.
That is good for population-level trend-spotting, and inadequate for diagnosing a sleep disorder.
Why deep sleep is the noisiest number
Deep sleep (slow-wave sleep, N3) is defined by delta-wave activity on EEG that no wrist device can measure. Wearables infer it from HRV troughs and motionlessness — but those signatures also occur during quiet wakefulness in highly conditioned athletes, leading to false positives.
If your device shows you got “2 hours of deep sleep” in a night, the 95% confidence interval is somewhere in the range of 60 to 160 minutes. That is not a precision instrument. It is a directional compass.
What to actually do
- Track totals, ignore individual nights. A single night’s “low deep sleep” reading means almost nothing. A 7-day moving average means something.
- Look at recovery scores in context. Whoop’s “recovery” is mostly HRV-relative-to-baseline. It is a useful signal — but it indexes your autonomic state, not your sleep quality per se.
- If your readings are persistently bad and you feel bad, talk to a clinician. Wearables are not a diagnostic instrument. A formal sleep assessment by a qualified physician is.
What we recommend reaching for first
Before any supplement: cool the room (65°F is the consensus target), no screens for the 60 minutes before bed, consistent wake time including weekends. These three interventions outperform any commercial sleep aid we have ever evaluated.
If you have run those for 30 days and still have trouble: 100–200 mg of magnesium glycinate 60 minutes before bed has the cleanest evidence base. We sell it. We also told you to try the room temperature thing first.
Premium supplements you can verify.
Seven formulations. Single-origin sourcing. Third-party tested by two independent labs.
Shop the line →
