S4. Persona Matching + Clustering (Momo)

📅Published 2026-05-14·🔄Updated 2026-06-30·⏱1 min read

1. Five Lifestyle Personas

Persona	Signals
Live impulse buyer	High LiveStreamPurchase share · High viewing time
Search explorer	Many SearchEvents · High cart abandonment
24-hour delivery dependent	Immediate OrderTransaction checkout · SLA-sensitive
Family shopper	Many household SKUs · Weekend purchases
TV home-shopping loyalist	High TVPurchase share · 50+ age group

2. Data Mix

• RFM (Customer)
• Live-viewing affinity (LiveStream × Viewer)
• Channel share (App / Web / TV / Live)
• Social personas (Dcard · Instagram keywords)

3. Output UI

• 5-persona card distribution
• KMeans 6-cluster (LLM labels)
• PCA scatter plot

4. Demo Scenarios

1. Single match → "Live impulse buyer 0.58"
2. 5K batch → 6 clusters
3. Cluster → live campaign (S5)