Retail Product Search: A Practical Approach at Target
Target says hybrid lexical-vector search lifted commerce metrics while cutting zero-result queries roughly in half.
The paper describes a production product-search system that combines keyword matching with vector retrieval, then controls precision before fusing result channels with weighted interleaving. Target reports online A/B gains of 0.97% in click-through rate, 0.98% in order conversion, and 1.10% in demand per visitor over lexical-only search. The authors say the system is deployed at scale and serves millions of guests daily. ArXiv · AI/CL/LG's note
The paper describes a production product-search system that combines keyword matching with vector retrieval, then controls precision before fusing result channels with weighted interleaving. Target reports online A/B gains of 0.97% in click-through rate, 0.98% in order conversion, and 1.10% in demand per visitor over lexical-only search. The authors say the system is deployed at scale and serves millions of guests daily. ArXiv · AI/CL/LG's note
score 4