Phase 1 — MVP (Months 0–2)
Event webhook, bulk intelligence scan, 5-stage lifecycle scoring, per-customer repurchase interval calculation, four intervention types, Brevo email delivery, Supabase persistence, Slack alerts, Lovable dashboard with live metrics.
Success gate: All stage transitions detected correctly. Intervention emails delivered within 60 seconds of trigger. Dashboard reflects real-time data after every scan.
Phase 2 — Intelligence Expansion (Months 3–4)
- Intervention outcome tracking — Log whether a customer's churn score improved within 14 days of receiving an intervention. Surface intervention success rate by type on the dashboard.
- Gemini-powered personalised email copy — Replace template-based email generation with a Gemini call that writes contextual copy based on the customer's purchase history, product affinity, and lifecycle stage.
- Product affinity signals — Enrich the scoring engine with which products a customer buys most frequently. Use this to personalise win-back offers.
- Multi-channel delivery — WhatsApp Business API integration for markets where WhatsApp open rates exceed 80%.
Success gate: Intervention outcome tracking live. At least one Gemini-generated email variant showing measurably higher reply rates than template version.
Phase 3 — Platform Intelligence (Months 5–6)
- Shopify native integration — Replace demo webhook with Shopify webhook listener firing on real order and refund events.
- Cohort analysis dashboard — Track which customer acquisition months produce highest-LTV cohorts.
- Predictive churn window — Train a lightweight model on intervention outcome data to predict churn probability from engagement signals beyond purchase frequency.
- White-label deployment — Multi-tenant SaaS packaging for D2C agencies managing multiple brands.
Success gate: First real Shopify store connected. White-label deployment live for one agency client.
Strategic Principle
Depth before breadth. Each phase builds on validated learnings from the previous stage. The platform phase only begins once the intelligence layer has proven reliable enough to be a client-facing product.