The Science Behind SleepAnalytics
Every metric we surface is derived from peer-reviewed sleep research. This page explains the science, the formulas, and the limitations — transparently.
Why Sleep Quality Matters Beyond Duration
7 hours of fragmented sleep is not the same as 7 hours of restorative sleep.
The American Academy of Sleep Medicine (AASM) defines sleep quality through multiple dimensions: Sleep Efficiency, Latency (time to fall asleep), Wake After Sleep Onset (WASO), and stage distribution — not just duration. The National Sleep Foundation's 2015 guidelines established age-based duration recommendations, but explicitly noted that quality metrics are equally important.
How Apple Watch Detects Sleep Stages
Sensor Fusion
Apple Watch uses three data streams simultaneously:
- Accelerometer: Detects micro-movements and breathing-associated motion
- Optical PPG: Heart rate and heart rate variability throughout the night
- Machine Learning: A neural network trained on polysomnography (PSG) validation data
The result is 4-stage classification: Core (N1+N2), Deep (N3), REM, and Awake.
Accuracy
Apple Watch's sleep staging has been independently validated. Key figures from Apple's dataset:
- Mean kappa: 0.63 (fair-to-moderate agreement with PSG)
- Sleep sensitivity: 97.9% (correctly identifies sleep)
- Wake specificity: ~75%
AASM Stage Mapping
| Clinical Stage | % of TST | Apple Watch Label | Why It Matters |
|---|---|---|---|
| N1 (light transition) | 2–5% | Core | Easily disrupted; brief transition state |
| N2 (light stable) | 45–55% | Core | K-complexes and sleep spindles; memory consolidation |
| N3 (deep / slow-wave) | 10–20% | Deep | Growth hormone release, physical recovery, glymphatic clearance |
| REM | 20–25% | REM | Memory consolidation, emotional regulation, creativity |
Our Sleep Score — Full Formula
We believe you should understand exactly how your sleep score is calculated.
The 8 components below add up to a maximum of 100. Each component uses published clinical thresholds — not proprietary weights.
| Component | Max Points | Threshold / Formula |
|---|---|---|
| Duration | 25 | NSF age-adjusted recommendation (±30 min earns partial score) |
| Sleep Efficiency | 20 | Linear: 75%→0 pts, 95%→20 pts |
| Sleep Latency | 10 | <10 min→10, 10–20→7, 20–30→4, >30→0 |
| WASO | 10 | <10 min→10, 10–20→7, 20–40→4, >40→0 |
| Deep Sleep % | 10 | 15–20% optimal; below or above reduces score proportionally |
| REM % | 10 | 20–25% optimal; below or above reduces score proportionally |
| HRV Trend | 10 | vs. personal 30-day rolling baseline (±10% earns partial score) |
| Sleep Regularity | 5 | SRI deviation from a 75-point target |
Inspired by: Pittsburgh Sleep Quality Index (Buysse et al. 1989) and NSF clinical thresholds (Ohayon et al. 2017).
NSF Age-Based Sleep Recommendations
| Age Group | Recommended Hours | Notes |
|---|---|---|
| Teenagers (14–17) | 8–10 h | |
| Young Adults (18–25) | 7–9 h | |
| Adults (26–64) | 7–9 h | Most App users fall here |
| Older Adults (65+) | 7–8 h | Architecture changes with age; less deep sleep is normal |
SleepAnalytics uses your configured age group to calibrate your duration score. Set it in Settings → Profile.
The Sleep Regularity Index (SRI)
The SRI, developed by Phillips et al. (2017), measures how consistent your sleep-wake timing is day to day. For each pair of consecutive days, it calculates the probability that you are in the same sleep state (asleep or awake) at every minute of the day.
An SRI of 100 means completely identical sleep patterns every day. An SRI of 0 means completely random timing.
Population median SRI: ~60. Athletes with consistent training schedules typically score 65–80.