Personal baselines, not population norms
Early Signal measures personal variations in overnight biometric baselines to suggest low-risk rest actions. It is a wellness tool, it does not diagnose or treat disease, and it is not medical advice.
01
Compared only to you
No population averages — the useful bit is a change from your usual pattern
02
Temperature and breathing lead
Those first. HRV, resting heart rate, and sleep back them up
03
Rare alerts by design
At most one Off from usual push every 3 days. A heads-up you can actually trust
How scoring works
01
Compared only to you
Early Signal builds a personal baseline from your own overnight biometrics. It never compares you to population averages or age-group norms. That matters because “normal” resting heart rate, HRV, and overnight temperature vary widely between people—and the useful signal is a change from your usual pattern, not a number on a chart.
02
Robust baselines (median and MAD)
Instead of a mean and standard deviation, the app uses a median and median absolute deviation (MAD). Outliers—one bad night after travel, a rare late workout—pull a mean around, but barely move a median. Robust z-scores keep day-to-day noise from looking like a crisis while still surfacing multi-day shifts.
03
Why temperature and breathing lead
Wrist temperature (sleeping) and respiratory rate are the strain differentiators among common Watch metrics. HRV, resting heart rate, and sleep still matter as supporting and contextual signals, but temperature and breathing get higher weight when they are present. Trends charts use the same order—temp and breathing first when available. If your Watch cannot provide temperature (pre–Series 8), Temp is omitted from the ring and a footnote notes that wrist temperature needs Series 8 or later. The app still scores the signals you have. Missing data never looks in range.
04
Weighted agreement, not a single spike
Morning status is Typical, Slightly off, or Off from usual. Off from usual generally needs agreement across weighted signals—or a sustained overnight temperature rise versus your own baseline. That design reduces alerts from one noisy metric alone. After a hard workout or a logged context, generic (non temperature-led) patterns can show as Explained by context instead of a push.
05
Rare alerts by design
Off from usual push notifications are capped (at most once every 3 days). Some lifestyle contexts can hold a push for generic patterns while still showing status in-app; temperature-led results are not held back by those contexts. Slightly off pushes are optional Pro. The goal is a heads-up you can trust, not a daily score to obsess over.
06
What the research backdrop is—and is not
Public research has explored how wearable biometrics can change when people feel unwell. That literature informs why temperature, breathing, and heart metrics are interesting—not a claim that Early Signal diagnoses, predicts, or names any disease. The app is a general wellness tool. Status levels describe personal deviations; they cannot tell you why a pattern occurred.
Background references
Public docs and overviews. Cited as context—not as clinical validation of Early Signal.
- Apple — Sleeping Wrist Temperature (Health)
- Apple — Respiratory Rate in Health
- Heart rate variability overview (NIH / MedlinePlus)
- Mishra et al., 2020 — Pre-symptomatic detection of COVID-19 from smartwatch data (Nature Biomedical Engineering). Background research, not a claim that Early Signal detects disease.
- Quer et al., 2021 — Wearable sensor data and self-reported symptoms for COVID-19 detection (Nature Medicine). Background research, not a diagnosis.
- Smarr et al., 2020 — Feasibility of continuous fever monitoring using wearable devices (Scientific Reports)
- Grzesiak et al., 2021 — Wearable sensors to detect influenza and the common cold before symptom onset (JAMA Network Open). Research context, not a claim that a watch names an illness.
Prefer product how-tos? See How it works, Guides, and our Editorial policy.
Science FAQ
Wellness scope, missing temperature, and why training can look like Slightly off.