Sign In Join Early Access

How It Works

From wrist signal to cycle intelligence in three steps

Your existing biosensor already captures the raw data. Clair Health's pipeline transforms temperature, heart rate variability, and skin conductance into calibrated cycle phase predictions with per-day confidence intervals.

Get Started Free
01

Connect your biosensor data

Clair Health is device-agnostic. Export a CSV from any supported wrist biosensor and upload it directly to your dashboard. No proprietary hardware required.

Wrist Biosensor Skin temperature Heart rate variability Skin conductance Activity index CSV Ingestion Layer Schema validation Timezone normalize Gap interpolation Artifact rejection Feature Engine Rolling HRV metrics Thermal delta (24h) Circadian alignment Conductance slope Clair Engine LSTM inference Phase classification Confidence scoring Calendar output

Skin temperature

Basal skin temperature captured at the wrist shows a reliable 0.3-0.5 C rise at ovulation and elevated luteal baseline that drops at menstruation onset.

Heart rate variability

HRV varies predictably across the cycle. Progesterone-dominant phases suppress parasympathetic tone, producing characteristic RMSSD depression we detect and model.

Skin conductance

Electrodermal activity patterns shift with hormonal milieu. Estrogen-dominant follicular phase produces distinct conductance slopes that our feature engine extracts nightly.

02

Calibrate to your personal baseline

Population norms are a starting point, not the destination. After 14 days of data, Clair Health builds a personalized physiological model anchored to your cycle's characteristic signatures.

Why personal calibration matters

Two people with identical cycle lengths can have thermal signatures that differ by 0.8 C. A model trained on aggregate data would misclassify both. We anchor to you, not the average.

  • 14-day warm-up period, no historical data needed
  • Recalibrates automatically if cycle length shifts
  • Identifies artifact windows and excludes illness noise
  • Works with irregular cycles, post-OCP recovery, perimenopause

Calibration progress

Day 1
Uploading
Day 7
Building
Day 14
Active

Personal model ready. Confidence intervals enabled.

03

Interpret your phase with confidence

Every day in your calendar carries not just a phase label but a confidence interval. You see how certain the model is, and your clinician can annotate deviations that warrant follow-up.

July 2025 Live cycle view
M1
M2
M3
M4
F5
F6
F7
F8
F9
O10
O11
O12
L13
L14
L15
L16
L17
L18
Ovulatory window
Day 11 of 28
Confidence
87%

Daily phase label

Each calendar day shows your predicted phase: menstrual, follicular, ovulatory, or luteal. Phase windows are model-inferred, not calendar-averaged.

Confidence intervals

Every day carries a confidence percentage. Low-confidence days are flagged for review rather than presented as certain. Uncertainty is data too.

Clinical export

Cycle+ and Clinical plans generate PDF phase reports with confidence band overlays your gynecologist or endocrinologist can annotate directly.

Architecture overview

A privacy-first, end-to-end pipeline from raw sensor bytes to structured phase output.

User Device CSV / wearable app encrypted upload Ingestion Schema check Timezone norm Artifact filter Gap interpolation Feature Engine Rolling HRV (24/48h) Thermal delta Circadian phase EDA slope LSTM Model Personal baseline Phase inference Confidence scoring Interval calibration Anomaly flagging Recalibration Output Dashboard PDF report API export All processing occurs on isolated compute. No raw sensor data retained after feature extraction. SOC 2 Type II in progress.

Common questions

  • Clair Health ingests CSV exports from any device that records skin temperature, HRV, and an activity index at least every 5 minutes. Confirmed compatible devices include Fitbit Sense, Oura Ring Gen 3, Garmin Vivosmart 5, Apple Watch Series 8 and later (via Health export), and Whoop 4.0. We publish the full schema specification so developers and OEMs can add compatibility.

  • The personalized model completes its warm-up calibration after 14 days of uploaded data. During that window you receive population-level phase estimates with wider confidence intervals. After calibration, accuracy and interval width both improve noticeably as the model adapts to your individual physiological signatures. Predictions continue to sharpen through the first three full cycles.

  • Yes. The LSTM architecture does not assume fixed cycle length. It learns your physiological rhythm directly from sensor data rather than fitting a 28-day calendar template. Users with PCOS, post-oral-contraceptive recovery, perimenopause, and cycle variability of plus or minus 10 days have shown successful calibration. Very short cycles (under 21 days) and anovulatory stretches produce lower confidence readings, which are honestly reported rather than suppressed.

  • Your raw sensor data is encrypted in transit and deleted from our servers within 24 hours of feature extraction. We store only derived feature vectors and model outputs, never personally identifiable raw waveforms. Clair Health does not sell, license, or share your data with third parties. Clinical users operate under HIPAA-aligned data handling agreements. Full details are in our Privacy Policy.

  • Clair Health provides cycle phase intelligence as a wellness and clinical-support tool. It is not cleared or approved as a contraceptive device. Ovulatory window predictions carry useful signal but are not sufficient as a sole contraceptive method. We actively recommend consulting a reproductive health clinician if you are making family planning decisions with this data.

Explore the science Start for free