Trailwise
Product intelligence that turns user behavior into testable experience recommendations.
Trailwise maps user journeys across websites and mobile apps, identifies behavioral signals associated with activation, conversion, and retention, and turns...
Role & Stack: Product strategy, UX design, analytics architecture, and implementation. Built with Next.js App Router, React, TypeScript, Tailwind CSS, Recharts, d3-sankey, Radix UI, Zod, and Prisma on SQLite, with server-sent events for live activity and Vitest for unit tests. Deployed on Vercel with deterministic synthetic demonstration data.
Overview: Analytics tools show events, funnels, journeys, and retention, but teams still have to decide what the behavior means, who is affected, what should change, which metric should improve, and which guardrail should be protected. Trailwise connects behavioral measurement, signal discovery, recommendations, and interactive experience previews in one workflow. It does not replace experimentation — it helps teams turn observed behavior into a clearer experiment hypothesis.
How It Works: Capture structured web and app events, resolve anonymous users, identified users, and sessions, calculate journeys, funnels, retention, and drop-offs, identify behavioral signals associated with a selected goal, generate an evidence-linked experiment recommendation, and preview the original and recommended journeys in Experience Studio. Measured analytics stay separate from product hypotheses.
Product Capabilities: Overview, live activity, Journey Explorer, funnels, user profiles, behavioral signals, recommendations, and Experience Studio. Two workspaces tell the same story twice: Aurelia Web covers website acquisition, desktop, tablet, and mobile-web behavior, signup and onboarding, wearable connection, practice-plan creation, friend invitations, and subscription conversion. Aurelia App covers iOS and Android behavior, mobile onboarding, permission decisions, first practice, reminder setup, return behavior, and paywall and trial conversion.
Experience Studio: Recommendations become experiences, not documents. Teams review the problem and supporting evidence, see the target segment, inspect the experiment hypothesis, compare original and recommended variants, interact with a website or mobile-app preview, start a synthetic tester session, generate events through interaction, compare the session with historic behavior, and monitor the primary metric and its guardrail.
Architecture: Web and app events flow into ingestion and validation, identity and session resolution, a unified event and user store, then journeys, funnels, retention, and signals, then recommendation rules, then Experience Studio. Ingestion validates payloads with Zod, redacts denylisted properties, and de-duplicates by eventId. Identity resolution merges anonymous visitors into known users with 30-minute sessions. Analytics and signals are computed from stored events with Prisma; recommendations come from a deterministic rules-and-scoring engine, not a trained model.
Trust: Analytics are calculated deterministically. Signals compare users with and without a behavior, keeping sample sizes and percentage-point differences visible, and confidence is bounded by interval width and a minimum sample threshold. Relationships are correlational, recommendations are experiment hypotheses, and estimated impact is not observed production lift. The public demo uses deterministic synthetic data with no real customer or personally identifiable information.
Outcome: Trailwise demonstrates how product analytics can move beyond reporting. It connects event data to user journeys, journeys to behavioral evidence, evidence to product recommendations, and recommendations to an experience a team can inspect and test.