🧭 The principle: you are your own benchmark
Every PauseCafé analysis rests on a simple idea: don't compare you to an average, compare you to yourself. This is known as an “N-of-1” approach (a sample of one: you), or self-tracking.
Why this choice? Because responses to caffeine vary enormously from person to person — how fast you clear it depends on genetics (the liver enzyme CYP1A2), smoking, certain medications, pregnancy and age. Telling you “the average person sleeps X hours” would mean nothing for you. Comparing your own days against each other, however, reveals a signal that belongs to you.
🎯 The golden rule: association ≠ causation
An analysis might show that your shortest nights coincide with your most caffeinated days. That does not prove that coffee shortened your sleep: other causes (stress, screens, a late meal) could explain both at once. PauseCafé shows you statistical coincidences in your data — leads to explore, never verdicts. This is a wellness app, not a medical device.
📊 The method shared by all the analyses
The three analyses (water, sleep, heart rate) share the same mechanics. Here they are, step by step:
- We rank your days (or nights) by caffeine level. Each day is assigned a total caffeine intake.
- We isolate the two extremes. The 25% most caffeinated days on one side, the 25% least caffeinated on the other.
- We compare the metric between those two groups. For example: on average, do you drink more or less water on heavily caffeinated days than on quiet ones?
- In addition, we compute a rank correlation coefficient (Spearman) from 7 days of data, to quantify the link across all your days, not just the extremes.
Why compare the extreme 25%, rather than all the days?
✅ A clear contrast, robust to “average” days
Middling days (neither heavily nor lightly caffeinated) dilute the signal: they look like everyone else's. By setting the top and bottom of the range against each other, we maximise the contrast and get a readable message (“on days you push the coffee, you sleep less”). It's also more robust : a single unusual day carries less weight when you reason in groups than in an overall average.
Why Spearman, and not Pearson?
✅ A rank correlation, suited to small, noisy datasets
Pearson's coefficient measures a linear relationship and assumes roughly normal data: it's sensitive to outliers (a single 5-litre day can throw everything off). Spearman works on ranks (how the days are ordered), not raw values. The result: it detects any monotonic relationship (“more coffee → less water”, even if it isn't a perfect straight line), it assumes no particular distribution, and it withstands outlying days. On human behaviour data — sparse and noisy — it's the cautious choice.
Why 7 days, then 14?
✅ Enough data points that it isn't just chance
Below 7 days, any result would be anecdotal: too few points, chance dominates. From 7 days/nights onwards, a trend starts to emerge — and PauseCafé says plainly that it's still fragile. From 14 onwards, the “top / bottom 25%” groups hold enough days to give a more reliable and actionable comparison. We'd rather say “hang on a little longer” than show a misleading figure.
💧 Analysis 1 — Water ↔ caffeine Premium
The first analysis cross-references your water intake and your caffeine, day after day. It answers one question: how does your hydration change depending on whether you drink a lot or a little coffee? Depending on the pattern, the message differs: “you over-compensate” (more water on caffeinated days), “you trade off”, or “steady hydration”.
📐 A worked example (over 14 days)
Suppose we isolate your extreme days:
Reading: on the days you push the coffee, you drink noticeably less water. Something to act on: remember to hydrate more on those days.
Limitations. This is your behavioural pattern, not a physiological law. Coffee didn't “dry out” your body: you simply drank less water on those days. The analysis describes; it doesn't explain.
😴 Analysis 2 — Caffeine ↔ sleep Premium
With your permission, PauseCafé reads your sleep duration from Apple Health (on your device only) and cross-references it with your caffeine. This analysis rests on solid scientific ground.
What the science says
A landmark study (Drake et al., 2013) gave 400 mg of caffeine to healthy sleepers 0, 3 or 6 hours before bed. The result: even taken 6 hours before sleep, caffeine cut sleep time by more than an hour, to a statistically significant degree. That's what justifies the app's “sleep threshold” (≈ 100 mg, the EFSA benchmark) and the A → G sleep impact grade: they estimate how much caffeine will be left when you go to bed.
What we measure
We compare your average night length against your caffeine intake, using the shared method: your most caffeinated nights (top 25%) against your least caffeinated nights (bottom 25%), from 7 measured nights (a trend), 14 for a more reliable comparison.
📐 A worked example
Reading: in your data, nights following heavy caffeine are on average around 55 minutes shorter. Worth watching — without concluding there's a single cause.
Limitations. Duration measured by a watch is not laboratory polysomnography. And other factors (stress, screens, alcohol, a late meal) affect sleep alongside caffeine: the coincidence observed does not pin the blame on coffee alone.
❤️ Analysis 3 — Caffeine ↔ resting heart rate Premium
Again with your permission and read on your device, PauseCafé can cross-reference your resting heart rate with your caffeine. This is the most cautious analysis — and we should be honest about the science.
What the science says (with nuance)
Caffeine stimulates the sympathetic nervous system (by blocking adenosine and releasing catecholamines). Reviews show a fairly clear acute effect on blood pressure, but less consistent data on heart rate. More importantly, tolerance builds in regular consumers: the body adapts, and the effect on resting heart rate becomes variable, sometimes small. That is precisely why we present it as a personal exploration, not as an established relationship.
What we measure
Same comparison method: your average resting heart rate on the most caffeinated days (top 25%) against the least caffeinated (bottom 25%), from 7 measured days, 14 for greater clarity.
📐 A worked example
Reading: a 2 bpm gap is small and could come from many other things (activity, hydration, stress). We present it as an observation, not as a strong signal.
Limitations. Caffeine tolerance, many confounding factors (exercise, sleep, hydration, stress), and a measurement from a consumer sensor. This analysis invites observation, never worry. If you have any doubt about your heart, consult a healthcare professional.
🧠 Coming soon — Caffeine ↔ stress Soon
An upcoming analysis will cross-reference your caffeine with a self-report questionnaire (stress, anxiety). The method will stay the same: comparing your most and least caffeinated periods, backed by a rank correlation.
Methodological honesty. Self-reported stress is subjective, and the link is probably bidirectional : stress drives you to drink more coffee, and coffee can heighten anxiety in some people. The analysis will help you spot your own pattern, without settling which way the arrow points.
🚫 What PauseCafé does not do
Rigour also means knowing what you refuse to do:
Claim a cause
We will never say “coffee reduced your sleep”. We show an association; it's yours to interpret.
Personalise half-life by body weight
Clearance depends on the liver (CYP1A2), genetics, smoking and medications — none of which the app can measure. Doing it would be false precision. Weight is used only where it's valid: the sleep threshold (≈ 1.4 mg/kg) and the daily goal (≈ 5.7 mg/kg).
Make a diagnosis
No analysis is medical advice. PauseCafé is a wellness app.
Send your health data anywhere
Sleep and heart rate are read from Apple Health on your device and never leave it: no health data is sent to our servers.
⚠️ Wellness information
PauseCafé's analyses are provided for guidance and have no medical value: neither a medical device, nor a diagnosis, nor medical advice. For any question about your sleep, your heart or your health, consult a healthcare professional.
Sources.
EFSA, Scientific Opinion on the safety of caffeine, 2015 (EFSA Journal 2015;13(5):4102) — average half-life ≈ 5 h, sleep impact threshold ≈ 100 mg, daily dose of no concern ≈ 400 mg (≈ 5.7 mg/kg) and ≈ 200 mg per single dose. Link →
Drake C, Roehrs T, Shambroom J, Roth T. Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed. J Clin Sleep Med 2013;9(11):1195–1200 — 12 healthy sleepers, 400 mg; significant effect up to 6 h before bed. Link →
James JE. The effects of caffeine on blood pressure and heart rate: a review (Annals of Behavioral Medicine) — marked acute effect on blood pressure, less consistent heart rate data, tolerance in regular consumers.
Nawrot P et al. Effects of caffeine on human health. Food Addit Contam 2003;20(1):1–30 — a synthesis of effects on sleep and the cardiovascular system.
Statistical method: Spearman rank correlation coefficient (a non-parametric measure of a monotonic relationship, robust to outliers).
Your insights, in PauseCafé Premium
Real-time active caffeine, sleep and heart rate correlations: leads to help you know yourself better, with your data respected.