User Profiling Across the Lifecycle: Applying AARRR and RARRA Models in E‑commerce

The talk by Yao Kaifei, co‑founder of Judo Technology, emphasizes that user profiling is essential for understanding customers and enabling business empowerment, especially in e‑commerce scenarios where profiles guide product‑chain optimization.

1. Understanding the two paths – The classic AARRR (Acquisition, Activation, Retention, Referral, Revenue) model provides a macro view of user behavior, while the RARRA model (Recognition, Channel selection, Personalized service, Re‑engagement, Revenue) offers a micro‑level framework for intervening at each lifecycle stage. Successful application requires both technical and business thinking.

2. Post‑visit analysis ("walked away") – By reversing the RARRA flow, teams can audit the full business chain, locate product issues, segment users (high‑value, churned, etc.), and use data from departed users to inform future acquisition and retention strategies.

3. Pre‑arrival ("before they come") – Existing user data guides channel and tag selection for new‑user acquisition; collaboration between AI, operations, and market experts helps choose the most effective advertising platforms and audience segments.

4. After arrival – Rapid feedback is needed for new users, especially to solve cold‑start problems. Users are split by strong vs. weak intent (e.g., specific product queries vs. generic terms), enabling tailored recommendations and content segregation while preserving the experience of existing users.

5. Review and optimization – Data‑driven testing (e.g., buying data for efficient experiments), global multi‑objective optimization, and closed‑loop/flywheel thinking create a continuous improvement cycle that aligns supply with demand, maximizes traffic distribution, and enhances profitability.

Conclusion – Key takeaways include: segment users by tags, avoid relying solely on averages, link user and product analyses, adapt dimensions to specific business scenarios, perform comparative analyses, and maintain a complete data loop to drive systematic optimization.