Skip to main content

S4. Persona Matching + Clustering (Mitsukoshi)

Published 2026-05-14Updated 2026-06-301 min read

1. URL · Personas

  • /personas · P2

2. 5 Lifestyle Personas

PersonaSignals
Luxury collectorRegular purchases of LV/Hermes/Chanel · Black tier
Seasonal trendsetterImmediate purchase of SS/FW new releases · Instagram
Foreign touristTaxRefund · FX-sensitive · single-visit purchases
Family shopperFood / weekends · average basket size
Accessory beginnerLow-price SKUs · first purchase

3. Data Mix

  • RFM (Customer + Foreigner)
  • Luxury affinity (purchase share of LV/Hermes/Chanel)
  • Xiaohongshu / Dcard keywords (social personas)

4. Output UI

  • Distribution of 5 persona cards
  • KMeans 6 clusters (LLM labeled)
  • PCA scatter plot

5. Demo Scenarios

  1. Single-record matching → "Luxury collector 0.62 / Seasonal trendsetter 0.18"
  2. Batch of 2,500 members → 6 clusters
  3. Clusters → campaign simulator (S5)