Launching today
Primrose & Eve: Menopause
Privacy-first self-advocacy tool for perimenopause.
10 followers
Privacy-first self-advocacy tool for perimenopause.
10 followers
Other perimenopause apps stop at symptom tracking. Primrose & Eve goes further: real statistical analysis (chi-square + FDR correction) on her data, and a clinically structured PDF report her doctor can act on in a 15-minute appointment. No central server, or trackers, or data sold. Health data lives on her iPhone and her own iCloud. Sign in with Apple, no email collected. We literally cannot see her data.
Mailwarm
How do you decide what makes it into the report without overwhelming the doctor with charts?
@naimz genuinely the hardest product call in the whole app.
Three layers of filtering happening before anything lands in the PDF: two on-device, and one through an AI summary layer with identity stripped before it leaves the phone.
Statistical significance gate (on-device). Our Pattern Engine runs Yates-corrected chi-square + Benjamini-Hochberg FDR correction (alpha = 0.10) on every trigger × symptom pair, locally. If it doesn't pass FDR, it doesn't make the report.
Effect size gate (on-device). Even if statistically significant, the risk ratio has to clear a meaningful threshold (≥ 1.5 for triggers, ≤ 0.6 for protective factors). "Real but tiny" gets cut here.
Hard editorial caps on the AI summary. Drivers are capped at 2-4, doctor questions at 3-5, strict word limits per item. The model is told to output fewer items if data is thin.
The output structure is also tuned for the 15-minute scan: symptom frequencies first, then trigger correlations, then targeted doctor questions. No raw chart dump.
Genuinely curious what you'd push on here. Anywhere you'd loosen or tighten this chain, or a gate I'm missing entirely?