Automate a Real Energy Forecast From Apple Watch Data, Not a Sleep Score
Why Apple's 0-100 Sleep Score cannot tell you when today's energy will peak or dip
What matters today
Why Apple's 0-100 Sleep Score cannot tell you when today's energy will peak or dip
Key points
- What the Sleep Score Actually Measures (and Stops At)
- The Two-Process Model, in Plain Language
- The Signal Sleep Debt Misses: HRV and Resting Heart Rate
- Build the Automation: About 25 Minutes, One-Time Setup
- The Prompt to Send
What You'll Learn
- Why Apple's 0-100 Sleep Score cannot tell you when today's energy will peak or dip
- The three signals that actually predict daily energy: sleep debt, circadian timing, and today's HRV and resting heart rate
- A step-by-step Shortcuts build (about 25 minutes) that pulls Health data every morning automatically
- The exact prompt to send your AI app for clock-time energy windows, not general advice
- Packaged apps like Welltory as a zero-setup alternative, and where the estimate still falls short
Say your Apple Watch handed you a Sleep Score of 84 this morning. Feels like a green light. So you book your hardest negotiation call for 2:30 p.m., the exact window when most people's cognitive performance quietly dips for the day. The score never warned you, because it was never built to.
Executives who schedule deep work off a single overnight number are gambling with hours they cannot get back. Under watchOS 26.2's revised thresholds, 5 to 6 hours of restless sleep can still land a "High" score, according to MacRumors. A good-looking number and a low-energy afternoon can both be true on the same day, and the gap costs real output: sleep scientists estimate cognitive performance swings roughly 5 to 20 percent from a person's daily peak to their daily trough, driven by a separate clock the Sleep Score was never designed to track.
5 to 20%
the swing in cognitive performance between your daily peak and your daily trough, a gap no Sleep Score is built to show you
What follows is a way to close that gap using only an iPhone, an Apple Watch, Apple Health, and Shortcuts. No new health app, no new hardware, no paid subscription required, one 25-minute build that turns recovery signals into a daily clock-time forecast every morning after.
What the Sleep Score Actually Measures (and Stops At)
Apple's Sleep Score, introduced in watchOS 26 and adjusted again in watchOS 26.2 in November 2025, grades exactly one night on a 0 to 100 scale, built from three weighted inputs according to Apple's own support documentation and MacRumors' reporting on the update:
- Duration: worth 50 of the 100 points
- Bedtime consistency: worth 30 of the 100 points
- Interruptions (time spent awake overnight): worth 20 of the 100 points
The November update also tightened the labels. Very Low now covers 0 to 40 (previously 0 to 29), Low covers 41 to 60, OK covers 61 to 80, High covers 81 to 95, and Very High (the renamed "Excellent" tier) covers 96 to 100. Apple built the new ranges around its Heart and Movement Study and guidance from the American Academy of Sleep Medicine, the National Sleep Foundation, and the World Sleep Society, per 9to5Mac's coverage of the change.
Tighter labels do not change what the score is built to answer. It is a backward-looking summary of last night. It says nothing about 10 a.m. versus 3 p.m. today, because duration, bedtime, and interruptions do not encode time-of-day information at all. Two people with an identical 84 could have completely different energy curves at 2 p.m., and the score has no way to tell you which one you are.
The Two-Process Model, in Plain Language
Sleep researchers explain daily energy with two separate systems working at once, known as the two-process model. Process S is sleep debt: pressure that builds the longer you have been awake and only clears when you sleep. Process C is circadian timing: a roughly 24-hour signal run by a clock in the brain (the suprachiasmatic nucleus) that pushes alertness up during the day and down at night, independent of how tired the debt-tracker says you should be.
Sleep debt is not last night's shortfall. It is cumulative, typically tracked over roughly the past 14 nights, with more recent nights weighted more heavily, because a single good night barely dents a two-week hole. Sleep debt is considered the single strongest predictor of how someone feels and performs day to day, which is exactly the number the nightly Sleep Score misses by only looking at one night.
Circadian timing runs on its own clock. Research on cognitive performance under controlled conditions shows output oscillating roughly 5 to 20 percent peak to trough across the day, independent of sleep debt. Peak performance capacity tends to land in the evening window (studies place it around 8 to 10 p.m.), with the deepest trough overnight (roughly 2 to 6 a.m.) and a well-documented secondary dip around mid-afternoon, close to 3 p.m. Combine a rising debt with a circadian trough and the same "High" Sleep Score morning can turn into a rough 2:30 p.m. call.
Illustrative Example Day
A sample forecast shape, not a fixed schedule. Your own clock-time output will vary.
7-8:30am
High energy
11am-12:30pm
High energy
2:30-4pm
Low energy
8-9:30pm
High energy
2-6am
Low energy
The Signal Sleep Debt Misses: HRV and Resting Heart Rate
Sleep debt and circadian timing predict the baseline pattern most days. Neither one can see how your body is actually coping today: a stressful week catching up with you, the first signs of getting sick, or last night's second glass of wine. A 14-day observational study of healthy adults found that higher heart rate variability, specifically RMSSD, tracked with better self-reported sleep and lower fatigue and stress the next day. A separate 2026 study in people recovering from post-COVID fatigue found HRV and daily activity patterns predicted next-day fatigue better than sleep metrics alone, evidence that this signal carries real weight even outside a sleep lab.
Apple Watch reports HRV using a different calculation, called SDNN, sampled automatically throughout the day and during Mindfulness sessions. SDNN and RMSSD are both legitimate, widely used ways to measure heart rate variability, but they are not the same number, so do not expect an Apple Watch HRV reading to match a Whoop or Oura reading from the same night. What matters is the trend against your own baseline, not the absolute number: a lower-than-usual HRV paired with an elevated resting heart rate is a real physiological signal that today is not a full-recovery day, regardless of what the Sleep Score said this morning.
This is the same calculation behind Garmin's Body Battery and Whoop's Recovery Score, which both combine HRV, resting heart rate, respiratory rate, and sleep into one proprietary daily number. Apple does not ship an equivalent. Its own Vitals app, updated in February 2026, shows heart rate, HRV, respiratory rate, wrist temperature, blood oxygen, and sleep duration side by side, each flagged against your personal baseline, but stops short of combining them into one score. That gap may be deliberate. Apple has generally favored showing individual signals over collapsing them into one proprietary number, and that restraint is a defensible design choice, not necessarily an oversight. Deliberate or not, it still leaves Executives without a same-day answer, which is the actual gap this automation closes: asking an AI model to do the same synthesis Whoop and Garmin charge a subscription for, using data Apple Watch already collects for free.
HRV
How recovered your nervous system is today
Resting Heart Rate
Whether your body is under more strain than usual
Respiratory Rate
An early flag for stress or getting sick
Build the Automation: About 25 Minutes, One-Time Setup
The Shortcuts app on iPhone can already read the exact inputs this model needs, straight out of Apple Health, with no third-party health app required. Here is the build.
Want to test the idea before building anything? Skip ahead to "The Prompt to Send," pull today's numbers from the Health app by hand, and paste them into your AI app once. If the forecast is useful, come back and automate the pull. If not, the only cost was one prompt.
Pull the health inputs (about 10 minutes)
Open Shortcuts, tap the plus sign for a new shortcut, and add five "Find Health Samples" actions (Shortcuts also lists this as "Get Health Sample" depending on iOS version). Configure one for Sleep Analysis over the last 24 hours to get wake time, one for Sleep Analysis over the last 7 days to build the sleep-total trend, one for Resting Heart Rate for today, one for Heart Rate Variability over the last 24 hours, and one for Respiratory Rate over the last 24 hours (this one only returns overnight values, since that is when Apple Watch measures it). Add a "Calculate Statistics" action set to Average right after the HRV pull, since Apple Watch takes multiple HRV readings a day and a single one can be noisy. Each action returns raw Health data your shortcut can reference downstream.
Format the data into text (about 5 minutes)
Add a "Text" action and use the Shortcuts variable picker to insert the wake time, the 7-night sleep totals, today's resting heart rate, today's average HRV, and last night's respiratory rate into a single readable block. This becomes the data payload the AI model reads, so keep labels explicit (for example, "Wake time: 6:42 AM" and "HRV: 38 ms" rather than bare numbers).
Run the forecast prompt (about 4 minutes)
Add a Shortcuts action that can process text, then paste the prompt below with the Step 2 data block appended underneath. If your Shortcuts setup does not have a direct text-processing action available, add "Copy to Clipboard" after Step 2 and paste the same prompt and data block wherever you normally analyze private notes. The important part is the data structure, not the brand of the tool reading it.
Act on the answer (about 2 minutes)
If you used the Claude app's "Ask Claude" action, pipe its text output into a "Show Notification" action, or an "Add New Calendar Events" action, so the forecast lands on your lock screen or blocks the two lowest-energy windows directly on your calendar. If you pasted the prompt into another app by hand, just read the response there and block those same two windows on your calendar yourself, a few seconds slower but nothing extra to build.
Set the trigger (about 1 minute)
In the Automation tab, create a Personal Automation set to a time each morning (for example, 15 minutes after typical wake time). If every step runs through the Claude app's action, turn off "Ask Before Running" and the whole chain fires without you opening anything. If you are pasting into another app by hand, leave "Ask Before Running" on, so the shortcut preps your data and opens the app for you, and all that is left is one paste.
Total build time runs close to 25 minutes for someone who has never opened Shortcuts before. Five health pulls take a little longer than three, but each one follows the same pattern. After that, it runs itself every morning if every step uses the Claude app's action, or opens your chosen AI app with the data ready to paste if it does not. Either way, the setup work only happens once.
The Prompt to Send
Send your AI app this exact prompt, with the Step 2 data block appended underneath it:
Using my wake time, my sleep totals for the last 7 nights, today's resting heart rate, today's average heart rate variability, and last night's respiratory rate, estimate today's 3 highest-energy 90-minute windows and 2 lowest-energy windows. Base this on sleep debt, circadian timing, and how well my body appears to be recovering today, not a single sleep score. If my resting heart rate or HRV looks meaningfully outside my normal range, flag that directly instead of quietly folding it into the forecast. If any input is missing, say which one instead of guessing. Give clock times, not general advice.
This works because it names all three mechanisms that matter (debt, timing, and today's recovery signals) and explicitly rules out the single-night score as the basis. It also forces clock times over vague phrasing like "later in the afternoon," which is what makes the output usable for calendar-blocking instead of just interesting to read.
The Zero-Setup Alternative: Packaged Apps
Executives who would rather not touch Shortcuts at all have packaged options. Several consumer apps, including Welltory and others, sync with Apple Health and estimate a similar sleep-debt-plus-timing forecast without building anything, usually shown as a scrollable energy curve for the day with a watch complication so peaks and dips show up on the wrist without opening the app.
Most of these apps charge an annual subscription with a short free trial. It is the faster path if the goal is seeing a forecast today rather than owning the automation. The tradeoff is a second subscription and a proprietary model instead of a transparent prompt you can inspect and edit yourself.
One Limitation to Keep in Mind
Neither the Shortcuts build nor a packaged app measures real-time energy. Both estimate a pattern from historical inputs (past sleep, wake time, resting heart rate) run through a fatigue model. If an unusually stressful call, a skipped lunch, or three double espressos hits mid-morning, the forecast will not see it, because none of those inputs feed the model. Treat the output as a planning tool for scheduling deep work and recovery time, not a live readout to check hour by hour.
Action Steps Summary
1. Stop scheduling off the Sleep Score alone. Treat it as one-night hygiene feedback, not a same-day energy forecast.
2. Build the Shortcuts automation once. Chain five Health-data pulls, a formatted text block, and a direct send to your AI app of choice, about 25 minutes total.
3. Use the exact verbatim prompt. It names sleep debt, circadian timing, and today's HRV and resting heart rate directly, and forces clock times, not vague advice.
4. Set the Personal Automation trigger. Run it a few minutes after typical wake time and turn off manual confirmation so it fires with zero morning effort.
5. Try a packaged app if you would rather skip the build entirely. Apps like Welltory and others use similar underlying logic (sleep debt plus circadian timing), usually with a short free trial.
6. Block the two lowest-energy windows on the calendar. Route admin work and email there, and protect the three highest-energy windows for the work that actually needs a sharp mind.
This information is for educational purposes only and is not medical advice. Please consult a qualified healthcare professional before making changes to your health routine.
Three deep dives. Four useful moves. One email worth opening.
PromptHacker turns the AI firehose into practical next steps for work, health, family, and everything time keeps trying to steal.