Contextual Ranker in Production: Session-Aware Similar Item Ranking for Lower-Funnel E-Commerce

Abstract

This talk presents Contextual Ranker, a production ranking system for eBay's ''Similar Items'' module on product detail pages. The problem sits at the lower funnel of e-commerce, where users are evaluating whether to purchase the current listing that they are viewing (the seed item ) or switch to an alternative. In this setting, recommendation quality depends on preserving strong relevance to the seed item while also adapting to short-term session intent. Traditional item-to-item rankers capture seed relevance but fail to capture session dynamics, while sequential recommenders model behavior but often dilute the seed item's role as the primary anchor. Contextual Ranker addresses this balance with a seed-anchored session architecture that encodes recently viewed items with a transformer, computes both session-aware and session-independent candidate relevance, and fuses these heterogeneous signals through a multi-head attention module. We will discuss the modeling choices, latency and scaling constraints, and evaluation methodology needed to deploy this design in production. The system serves 10K+ QPS at 29\,ms p99 latency on CPU-only infrastructure, improves offline purchase NDCG@4 by 3.7%, and lifts purchase-through rate by 1.5%--2.4% in online experiments.