About
A small research project on a badly under-measured condition.
Endometriosis affects roughly one in ten people of reproductive age, yet day-to-day symptom data is rarely collected in a way that supports honest statistical work. HerPattern exists to close part of that gap: structured daily tracking, transparent forecasting, and research tooling that treats participants as collaborators.
Patterns, never verdicts
Every output is a probability with an uncertainty range. We do not diagnose, detect flares, or suggest treatment, and we say so on every screen that shows a forecast.
Methods in the open
The model card, evaluation protocol, and benchmark results are public. Statistical modelling runs in a separate service so results can be reproduced and challenged.
Participants stay in control
Consent is recorded per purpose, research exports are pseudonymous, and account deletion removes health records rather than hiding them.
How the project grew
Phase 1
Tracking foundation
Personalized symptom sets, manual and voice check-ins, and a daily burden score computed from what each person actually chose to track.
Phase 2
Forecasting service
Hierarchical Bayesian autoregression and a switching state-space model delivered by an external Python service, with calibration reported openly.
Phase 3
Research tooling
Study setup, compatible-study matching, pseudonymous cohort exports, and audit trails for research teams working with consenting participants.
Now
Pilot and evaluation
Running structured pilots with participants and research groups to test calibration, usability, and the value of pattern summaries in real routines.
Who works on it
HerPattern is built by a small, distributed team working across applied statistics, product engineering, and participant research, with input from people living with endometriosis. We do not publish personal contact details or office locations. Everything reaches us through the enquiry routes on the contact page.