# Design an E-Commerce Recommendation System from Browsing History Design a large-scale recommendation system that uses users' browsing histories to rank products. ### Constraints & Assumptions - Recommendations must respect availability and policy at serving time. - Impression data is required to interpret clicks. ### Clarifying Questions to Ask - Which surface and business metric matter? - What latency, freshness, and cold-start requirements apply? ### What a Strong Answer Covers - Event collection, candidates, ranking, features, feedback bias, evaluation, and serving. - Fresh inventory filtering, fallbacks, diversity, and monitoring. ### Follow-up Questions - How would anonymous sessions work? - How would exploration be introduced safely?