RetentionHealth
A system that detects patient drop-off risk without relying on personal data
Most systems are built around individuals. They track identities. They store personal data. But they don't define how risk actually emerges.
The Problem
Most systems are built around individuals. They track identities. They store personal data. But they don't define how risk actually emerges. In practice, patient drop-off isn't an individual event. It's a pattern. Signals appear early: engagement drops concerns increase But they are fragmented across systems. And when everything is tied to identity, detection becomes: slow noisy and expensive to manage from a compliance standpoint
The System
We didn't build a tracking system. We defined a signal system. A cohort-based model that detects risk without relying on personal data or identity. Instead of asking: "Who is at risk?" The system asks: "What patterns indicate risk?" No identity tracking. No personal data storage. Just structured signals and defined evaluation.
How It Works
EVENTS
SIGNALS
AGGREGATION
SCORING
RISK
EVENTS
Behavioral activity is captured as raw input
SIGNALS
Meaningful patterns are extracted
AGGREGATION
Signals are grouped into cohorts
SCORING
Deterministic evaluation of risk patterns
RISK
Drop-off risk is identified at the system level
System Architecture
The system is built on three principles: No identity dependency Risk detection does not rely on personal identifiers Signal-based modeling Behavior is evaluated through structured signals Deterministic evaluation Risk is computed—not inferred Everything runs without storing personal data.
Why This Matters
Before: Risk was tied to individuals. Detection was fragmented and reactive. After: Risk is defined at the system level. Patterns are detected early and consistently. — Visibility Clinics can see risk before patients drop off Speed Intervention happens earlier Compliance No personal data required — Result: Better retention without increased regulatory burden