CEP · Customer Engagement Platform

CEP Platform — Full Design & Data Flow

① Segment Builder feeds ②③④ directly (they are parallel — ② does not drive ③④). Business lines enter three ways: A growth funnel · C in-app recommendation · B direct to Delivery. Behavior flows back through Analytics.

Business Lines 💰 Botim Money 🧩 Mini Program 💳 Credit Card + more Mode C · In-app recommendation Mode A · Growth funnel (full pipeline) Mode B · Direct to Delivery (OTP / txn) CEP Platform 🗄️ Customer Data Platform 1 Segment Builder Defines WHO — outputs reusable audience segments feeds ②③④ directly 3 Personalize Experience Standalone · directly callable On-device · INBOUND (when the user opens the app) · Recommendations · Home banner · Feed · Cards · Pop-ups · Tab personalization Two entries: ① builder-direct ② direct call 2 Campaign & Journey [CEP] the orchestration brain · directly callable Decides WHEN / HOW to engage · One-off campaigns · Multi-step journeys (trigger · wait · branch) · A/B · frequency cap 4 Delivery [CEP] Standalone · OUTBOUND · directly callable Multi-channel · reach out to users SenderID + templates 💬 Botim msg 📱 SMS 🔔 Push ✉️ Email 📞 Open Call (smart call) Two entries: ① builder-direct ② direct call 📊 Analytics & Insight data segments — builder-direct to ②③④ request in-app recommendation directly direct to Delivery — fire on call (OTP / txn) behavioral feedback — refine audiences & strategy
Funnel & builder-direct flow Mode C → Personalize Mode B → Delivery Feedback loop
PHASE 2 · 二期

推荐系统 & 搜索优化 — Recommendation & Ranking Engine

二期新增一个共享排序智能层。它消费一期的 ① Segments、画像/内容数据与实时行为,统一驱动两个 surface:③ Personalize 的推荐位(复用一期模块)和一个新的 Search surface(搜索结果优化)。行为回流用于模型训练。

① Segments from Phase 1 · audiences 🗄️ Profile & Content items · MP · content meta 📈 Realtime Behavior clicks · dwell · convert Recommendation & Ranking Engine NEW · 二期 · shared intelligence layer Candidate generation → ranking models → online serving Feature store · embeddings · real-time features Exploration · diversity / freshness · A/B & guardrails 3 Personalize — Recommendations reuses existing ③ · INBOUND · Home “For You” / Feed ranking · Content & Mini-Program reco · Cards · related items 🔍 Search — Result Optimization NEW surface · query-driven · Query understanding + intent · Personalized result ranking · Suggestions · zero-result fallback 📊 Analytics — CTR · conversion · search success rate signals in serves recos ranks results behavior → model training (online learning)
Recommendation & Ranking Engine (new) Search surface (new) ③ Personalize (reused) Engine serves surfaces Behavior → training