Operating rules
This is a personal habit experiment, not medical advice. It is built to answer a narrow question: which repeatable behavior changes my own trend? It does not diagnose sleep or cardiovascular problems. If a red flag appears, stop the experiment and seek appropriate medical advice.
Trends, not individual days
Use a rolling 7-valid-row median. A single night or single RHR reading is noise unless it comes with symptoms or an alert.
One new variable at a time
Every rung adds one behavior. Successful earlier behaviors remain active. Week 6 of Mission A forces you to choose alcohol timing or fluid timing, never both together.
Adherence gate
Daily behaviors count only at ≥6 of 7 days. If a rung is scheduled fewer times (for example, 3 aerobic sessions), complete the full planned dose while retained daily behaviors still meet their own ≥6/7 gate. Otherwise repeat; do not advance.
Exclude broken observations
Use the row exclusion checkbox for illness/fever, travel or time-zone change, tracker not worn, sync failure, hand-edited sleep, abnormal events, or an unusually short/nonrepresentative primary sleep. Alcohol is usually a confounder, but do not exclude it when alcohol timing is the variable you are deliberately testing.
Personal baseline wins
Do not chase an arbitrary universal score or bpm. Google itself says to focus on your own Sleep Score trends rather than a perfect 100. Define success versus your baseline before each mission.
Keep the instrument stable
Use the same tracker, similar fit, consistent overnight wear, and morning sync. A new device, sensor, or algorithm can shift measurements and may justify a fresh baseline.
The experimental loop
BASELINE → ADD 1 VARIABLE → MEASURE → HOLD → KEEP / REJECT → ADD NEXT VARIABLE
The protocol is intentionally boring. Boring is good: if five things change at once, the data becomes decorative fiction.
Baseline week — change nothing. Record everything.
without a stable comparison, later improvement is a story rather than an experiment.
For seven valid days, intentionally change nothing. Wear the tracker consistently, sync each morning, and record the values below. If a day is excluded, use + replacement day until you have seven valid observations.
| Day | Date | Sleep score | Duration h:mm | Time to sound min | Sound sleep h:mm | Restless min | Full awaken. # | Interruptions min | RHR bpm | Bedtime | Wake | Exclude / reason |
|---|
Improve Google Health Sleep Score
sleep responds night by night, so it is the first system to stabilize before you deliberately add cardiovascular training load.
Current 2026 score model
Verified Google’s current Sleep Score is a personalized score based on several metrics covering falling asleep, continuity, and amount/quality of sleep. Google does not publish the weighting of each component.
| Metric | Google measures | Better direction |
|---|---|---|
| Sleep duration | Estimated sleep during the primary sleep window. | ↑ longer, within your actual need |
| Time to sound sleep | From first trying to sleep until Deep, REM, or steady Light sleep with a calm heart rate. | ↓ shorter |
| Sound sleep | Total steady, undisturbed Light/Deep/REM sleep with low, steady heart rate. | ↑ more |
| Restlessness | Brief wake-like or transitional moments; longer memorable awakenings are excluded. | ↓ less |
| Full awakenings | Distinct wake periods longer than 5 minutes. | ↓ fewer |
| Interruptions | Total time fully awake in longer (>5 minute) stretches within the sleep period. | ↓ less |
Verified Google bands: Excellent 90–100 · Good 80–89 · Fair 60–79 · Poor <60; most users average 72–83. Google explicitly advises focusing on personal trends rather than chasing 100.
Verified Naps ≥20 min are tracked. Google’s July 2026 release notes say naps now contribute to 24-hour total sleep duration. Not verified Google does not document naps as an input to Sleep Score; the score page defines its duration component using the primary sleep window.
The experimental ladder
-
Adds · new variable
Fixed wake time
wake timeRule: use the same wake target every day, weekends included; aim to get up within roughly ±15–30 minutes. If you set a Google Health wake target, the app awards its schedule star within 15 minutes of target.
Verified Google and NHLBI both recommend a regular sleep-wake schedule, including days off. My recommendation Wake time comes first because it gives the later sleep-window experiment a fixed anchor.
gate: held ≥6/7 days -
Adds · new variable
Protected sleep window
wake timesleep windowRule: work backward from your fixed wake target and reserve enough time in bed to support at least about 7 hours of actual sleep. Give yourself realistic settling time; do not count phone time in bed as sleep opportunity.
Verified CDC recommends ≥7 hours for adults 18–60, and Google says adequate sleep opportunity and the sleep-duration goal matter to the score. My recommendation Use your baseline time-to-sound median as the first buffer rather than inventing a universal 30-minute latency.
gate: held ≥6/7 days -
Adds · new variable
Caffeine cut-off
wake timesleep windowcaffeine cutoffRule: keep your normal morning caffeine amount unchanged. Change timing only. Chosen cut-off:
Verified NHLBI says caffeine can interfere with sleep and its effects can last up to 8 hours; Google lists caffeine as a factor that can raise heart rate. Reasonable inference We map this primarily to time-to-sound and sound sleep; Google does not prescribe that mapping.
gate: held ≥6/7 days -
Adds · new variable
30-minute repeatable wind-down
wake timesleep windowcaffeine cutoffwind-downRule: start a fixed 30-minute routine before lights-out. Keep it boring and repeatable: dim lights, screens off, shower, paper book, gentle stretching — choose a sequence you can actually repeat.
Verified NHLBI recommends quiet time before bed and limiting bright artificial light/screens. Reasonable inference This rung is mapped to time-to-sound sleep because that metric describes the settling phase.
gate: held ≥6/7 days -
Adds · new variable
Bedroom environment
wake timesleep windowcaffeine cutoffwind-downcool · dark · quietRule: make the room cool, dark, and quiet; place the phone away from the bed so notifications and clock-checking cannot become part of the sleep environment. No shopping spree required.
Verified NHLBI recommends a quiet, cool, dark bedroom. Reasonable inference We map this mainly to restlessness and interruptions; Google does not say a specific room intervention moves a specific sub-metric.
gate: held ≥6/7 days -
Adds · choose exactly one variable
Alcohol timing OR large-fluid timing
wake timesleep windowcaffeine cutoffwind-downroomone timing testChoose: Cut-off / rule:
Verified NHLBI advises avoiding alcohol before bed; Google lists alcohol as a factor that can raise RHR. Reasonable inference Large-fluid timing is an experimental lever for awakenings/interruptions, not a Google recommendation. Run the second timing variable as a repeat Week 6 if you want to test it.
gate: held ≥6/7 days -
Adds nothing
Hold and confirm
wake timesleep windowcaffeine cutoffwind-downroomaccepted week-6 testRule: two full weeks with no new variable. Compare both 7-valid-row medians with baseline. Keep interventions that remain practical and appear helpful; reject or retest ones that did not produce a stable signal.
My recommendation Two hold weeks are a confirmation gate, not a medical threshold. They reduce the chance that one unusually good week becomes a permanent rule.
required before Mission B
How to measure — Mission A
- Path in app
- Google Health → Sleep tab → Sleep duration → Score
Use the “Sleep duration” / “Sleep quality” chips for score components. Sleeping HR: Today → Sleep duration → More about your sleep → Sleeping heart rate. - When to read
- Every morning after sync. Google notes that the Sleep Score can take a minute to appear after syncing.
- What to record
- Score, all six score metrics, bedtime, wake time, retained rungs, and any exclusion reason. Log naps in notes when they might affect how rested you feel.
- What counts as noise
- A single night; a night with illness, travel, bad tracker wear, sync failure, or a hand-edited sleep log. Google warns edited logs may recalculate a score that is not fully accurate. An algorithm/device change can justify re-baselining.
- What counts as success
- Your predefined success criterion, met across both hold weeks, with no obvious subjective deterioration. Default protocol rule: the median Sleep Score beats baseline in both hold weeks and the metric targeted by the strongest rung moves in the expected direction.
Metric → lever map
this is the debugging table: if a metric stalls, you know which experiment to revisit rather than randomly adding habits.
| Sleep metric | Experiment | Week | Expected signal | Evidence status |
|---|---|---|---|---|
| Sleep duration | Protected sleep window | 2 | Duration median ↑ | Inference |
| Time to sound sleep | Wind-down | 4 | Minutes ↓ | Inference |
| Sound sleep | Caffeine timing | 3 | Sound-sleep duration ↑ while total duration is similar | Inference |
| Restlessness | Bedroom environment | 5 | Restless minutes ↓ | Inference |
| Full awakenings | Alcohol OR fluid timing | 6 | Count ↓ | Inference |
| Interruptions | Alcohol OR fluid timing | 6 | Minutes ↓ | Inference |
| Schedule consistency | Fixed wake time | 1 | Wake-time spread narrows | Protocol |
Reasonable inference Google documents the metric definitions and healthy-sleep principles, but it does not state that each behavior above maps one-to-one to a specific score component. Treat the table as an experimental hypothesis map.
Daily log — sleep
| Wk | Day | Date | Score | Duration h:mm | To sound min | Sound h:mm | Restless min | Awaken. # | Interr. min | Bed | Wake | Rungs kept | Exclude / notes |
|---|
Lower resting heart rate
begin only after Mission A’s hold weeks. Sleep is now part of the stable background; training becomes the new variable.
Re-baseline before training
My recommendation The original baseline is stale after Mission A because your sleep routine has changed. Use the RHR observations from the two sleep hold weeks as the new comparison period.
What Google currently documents about RHR
Verified Resting heart rate is the number of beats per minute when you are still and well-rested. Google says RHR is estimated each day; sleeping heart rate is detected continuously during the night and is often lower.
Verified Current app path: Health → Key metrics → Resting heart rate. Google lists stress, alcohol/caffeine, and fever as factors that usually raise RHR; regular exercise or meditation can lower it; temperature and some medications can also affect it.
Not verified Google’s public help page does not publish the exact current RHR derivation or weighting. It directs users to the in-app About resting heart rate → Learn more. This notebook does not infer the hidden algorithm.
The experimental ladder
-
Adds · new variable
30-minute easy walk
sleep protocoleasy walkRule: 30 minutes daily at an easy conversational pace. This is intentionally below “training heroics”; the goal is to establish repeatable movement with low recovery cost.
Verified CDC says some physical activity is better than none, and regular activity supports health. My recommendation Start below the eventual aerobic dose so the next rung can test a clear increase in intensity.
gate: held ≥6/7 days -
Adds · new variable
3 × 30–40 min Moderate-zone sessions
sleep protocoldaily movement3× moderateRule: upgrade three of the week’s walking blocks to 30–40 minutes in Google’s Moderate zone; do not stack the moderate session on top of the same day’s easy walk. Brisk walking, cycling, or easy jogging are fine.
Verified Google’s Fitbit zones define Moderate as 40–59% of heart-rate reserve and describe it as a pace where breathing/HR are elevated but conversation is still possible. CDC’s adult guideline is 150 minutes moderate/week (or equivalent); this rung deliberately builds toward that dose rather than jumping there at once.
gate: 3/3 sessions + retained daily habits ≥6/7 -
Adds · recovery variable
10-minute slow-breathing practice
sleep protocoldaily movement3× moderateslow breathingRule: 10 minutes at the same time each day using comfortable slow diaphragmatic breathing. Do not force breath-holds. Stop if you feel dizzy or light-headed.
Verified Google says meditation can lower RHR; AHA says slow/deep breathing can support calm and cardiovascular responses. My recommendation Ten minutes is the experiment dose. It is not a guarantee that RHR will fall.
gate: held ≥6/7 days -
Adds · volume variable
Make one aerobic session longer
sleep protocoldaily movement3× moderatebreathing1× longerRule: extend one Moderate session to about 60 minutes; keep the other two at 30–40 minutes and keep intensity similar.
My recommendation This isolates volume from intensity. You are changing session duration, not simultaneously making the workout harder.
gate: planned sessions completed; sleep not worse -
Adds · optional intensity variable
Replace one Moderate session with Vigorous work
sleep protocoldaily movement2× moderate incl. longbreathing1× vigorousOnly if: previous weeks were well tolerated, RHR trend has not worsened, and sleep/recovery remain stable. Replace one Moderate session with 20–30 minutes of Vigorous-zone work; do not simply add unlimited exercise.
Verified Google defines Vigorous as 60–84% of heart-rate reserve. My recommendation Intensity is last because it has the highest recovery cost and is unnecessary if earlier rungs already produce a useful trend.
optional rung -
Adds nothing
Hold and confirm
Rule: keep the accepted training/recovery dose unchanged for two weeks. Compare both RHR medians with the re-baseline and make sure Sleep Score did not deteriorate.
My recommendation Success is a sustained personal improvement without buying it with worse sleep or poor recovery. Your predefined “meaningful improvement” is the threshold; this protocol does not invent a universal bpm target.
final confirmation gate
How to measure — Mission B
- Path in app
- Google Health → Health → Key metrics → Resting heart rate
Sleeping HR: Today → Sleep duration → More about your sleep → Sleeping heart rate. - When to read
- Once each morning after sync. Google says RHR is estimated each day; do not treat repeated refreshes as separate measurements.
- What to record
- RHR, sleeping HR, Sleep Score, workout type/duration, Moderate/Vigorous minutes, recovery practice, active rungs, and exclusions.
- What counts as noise
- A single day; illness/fever, alcohol/caffeine outliers, extreme heat, poor device fit, travel, or an unusually hard session. Google notes physiology, placement, activity, temperature, stress, alcohol/caffeine, illness, and medication can affect heart-rate data.
- What counts as success
- Both hold-week 7-valid-row RHR medians beat your re-baseline by the amount you defined beforehand, while sleeping HR/Sleep Score and subjective recovery do not show a meaningful deterioration.
Daily log — resting heart rate
| Wk | Day | Date | RHR bpm | Sleeping HR | Sleep score | Workout | Workout min | Moderate min | Vigorous min | Breathing ✓ | Active rungs | Exclude / notes |
|---|
When this stops being a habit problem
This is where the protocol deliberately hands the problem to a medical professional. The items below are prompts to seek evaluation, not diagnoses.
- Loud snoring with gasping/choking, or witnessed breathing pauses. NHLBI lists loud snoring, breathing that starts/stops, and gasping as sleep-apnea symptoms.
- Persistent excessive daytime sleepiness, especially if it affects driving, work, concentration, or you fall asleep unintentionally.
- Persistent insomnia. If sleep difficulty is affecting daily life, discuss it with a clinician; NHLBI defines chronic insomnia as difficulty ≥3 nights/week for ≥3 months.
- Unexplained RHR trend rising week over week after obvious confounders are removed, or a sustained departure from your usual range.
- High/low heart-rate alert while inactive. Google Fitbit Air can alert when HR is outside thresholds while you appear inactive for at least 10 minutes; review the alert and share relevant symptoms/data with a clinician.
- Palpitations with chest discomfort, dizziness, fainting/near-fainting, or unusual shortness of breath. Seek medical evaluation rather than continuing the experiment.
- New or unusual breathlessness during exercise, exercise intolerance, or symptoms that make you stop a session unexpectedly.
- The numbers themselves are making sleep worse. If score-checking creates anxiety or compulsive behavior, stop optimizing the score and prioritize how you feel and appropriate clinical advice.
Local data controls
Everything entered here is stored only in this browser’s localStorage. There is no account, server, analytics endpoint, or network write.