Personal sleep analytics: find what actually affects your sleep

Measurements aren't answers

Most sleep apps show you what happened: hours slept, stages, a score. That's useful, but it leaves the most important question open — what can I do differently?

General advice helps only up to a point. "Don't eat late", "no screens", "no coffee after noon" are true on average. For you, one of them may matter a lot and another barely at all. Personal sleep analytics is about finding out which is which.

The idea: compare your nights with your nights

Research usually compares groups of people. Within-person analysis — also called N-of-1 — compares one person's nights with each other:

• nights after an early dinner vs. nights after a late one,
• nights with a drink vs. nights without,
• nights after an evening workout vs. rest days.

Because it's always the same person, many things that differ between people — age, genetics, fitness, the watch — stay the same. What's left is how your body responds.

What SAYA compares

Your evening (inputs): a quick evening log plus data from your Garmin, for example:

• dinner timing and a dinner score (timing, size, processing level)
• alcohol and coffee
• training timing, activity and steps
• evening stress and daily stress
• wind-down routines such as reading, breathing exercises or stretching
• things that keep you wired, such as late work, social media or an argument
• bedtime regularity and your evening resting heart rate

Your night (outcomes): Sleep Score, Recovery Score, readiness, deep and REM sleep, overnight HRV, HRV-Incline and morning resting heart rate.

Automatic metrics from your Garmin can be analysed from the first days. Habit correlations need logged evenings — the more you log, the more SAYA can learn.

What a result looks like

For each habit and outcome, SAYA shows the evidence side by side — for example:

Dinner Score → Sleep Score: r = 0.40 · top vs. bottom quarter +17% · p < 0.05 · 36 nights

• r — how closely the two go together (from −1 to +1).
• Effect — how much the outcome differs between high and low habit values.
• p — how likely a pattern this strong would be if there were no real relationship.
• Nights — how much data the result is based on.

Positive drivers and negative impactors are ranked, so you can focus on the habits that matter most in your data. How these numbers are calculated — and their limits — is described on the methodology page.

Association is not causation

A correlation in your data is a strong hint, not proof. Some typical reasons a pattern can mislead:

• Habits come in bundles. Late dinners often come with social evenings, drinks and later bedtimes. SAYA sees each habit separately, so it can't always tell which one really matters.
• The direction can be reversed. A stressful day can lead to both a late workout and a bad night.
• Chance. SAYA tests many habit–outcome pairs. Some results will look meaningful purely by chance, especially with few nights.

From pattern to personal experiment

The best way to check a pattern is to test it:

1. Pick one habit SAYA flags — for example, dinner timing.
2. Change only that habit for two to three weeks.
3. Keep the rest as stable as you reasonably can.
4. Compare your nights before and after.

If the difference holds up, you've found a lever that works for you. If it doesn't, you've saved yourself a rule you didn't need.

SAYA is a wellness product, not a medical device. Nothing here is a diagnosis or medical advice — if you suspect a sleep or heart condition, please talk to a professional.