Sleep tracking apps — what's actually accurate, what's noise

Sleep trackers — Fitbits, Apple Watches, Oura rings, phone apps — have made sleep data feel concrete. Most of what they measure is genuinely useful. Some of what they claim to measure is fictional. Knowing which is which prevents wasted optimisation.

What trackers measure accurately

Total sleep time (within 30 minutes). Movement-based actigraphy is reasonably good at distinguishing "in bed and still" from "in bed and moving" or "out of bed." For most adults, the total-sleep-time number is within 15–30 minutes of truth.

Sleep timing. When you went to bed, when you got up, when you woke during the night. These are very accurate.

Heart rate during sleep. Modern wearables measure resting heart rate well. Tracking it across weeks is a useful health signal independent of sleep specifically.

Movement during sleep. Restlessness is genuinely measured.

What trackers measure poorly

Sleep stages (light, deep, REM). This is where the marketing exceeds the science. True sleep staging requires EEG. Wrist trackers infer it from heart rate variability and movement, which correlate weakly with actual EEG. A tracker that says you got "1h 23min of REM" is making an educated guess accurate to about 60–70%. The number changes with the device. The number changes with the firmware update. Don't optimise it.

"Sleep score." A composite metric, weighted differently by each manufacturer, of dubious clinical meaning. Useful only as relative trend ("sleeping worse this week than last") — not as an absolute value.

Snoring detection (phone apps). Phone microphones in a quiet room are reasonably good. In a room with a partner, fan, or window noise — much less so.

Blood-oxygen levels. Modern wearables (Apple Watch, Oura, Fitbit Sense) include SpO2. Accuracy is moderate at best — fine for trend-spotting, not fine for diagnosing sleep apnoea, which still requires a medical sleep study.

What to actually do with tracker data

Track for 1–2 weeks, then stop. You'll learn your patterns: typical bedtime, average sleep duration, weekday vs weekend gap. After that, the data is repetitive and the daily checking becomes its own anxiety driver.

Watch trends, not single nights. A bad night happens. A bad week is a signal.

Use it to spot specific problems. If you discover you regularly wake at 3am, or that weekend lie-ins are 3+ hours longer than weekday wake times, that's actionable. The detail of which sleep stage you're missing is mostly not.

Don't optimise the score. People who chase higher sleep scores often go to bed earlier than they should, anxious about the number. The number is a proxy. Your subjective rest, energy, and mood are the actual outcomes.

The orthosomnia trap

Orthosomnia — anxiety about sleep tracking that itself disturbs sleep — is a real and growing pattern. It looks like: checking your sleep score before getting out of bed, feeling anxious if it's lower than expected, planning the next night around recovering the score. The score becomes the source of the problem it claims to solve.

The fix is to stop checking. Most people who have used a tracker for a year would benefit more from a month off it than from another year of data.

A simpler alternative

A spiral notebook on the nightstand. Each morning, write three things:

1. Approximate sleep time (in bed and lights out → wake)
2. Subjective restedness (1–10)
3. Anything notable (drank, late meal, alarm, woke at 3am)

Two weeks of this gives you 80% of what a tracker provides, with none of the gamification. After two weeks, stop. Trust your body's signals; they were working before the wrist sensor existed.

When tracking does help

For specific use cases, trackers genuinely help:

• Diagnosing irregular sleep patterns for clinical investigation (bring the data to a sleep clinic)
• Verifying suspected apnoea before requesting a sleep study (low SpO2 + many wakeups + loud snoring = ask your doctor)
• Confirming travel/jet-lag recovery patterns so you can adjust the Jet-Lag Planner you use next time

Outside these, the data is mostly an interesting distraction. The framework's view: instruments measure; bodies feel. Trust both, but more the latter.