Case Study: Vipshop achieves 10x faster recommendations with Zilliz Milvus

VIPSHOP Builds a 10x Faster Personalized Recommender System Using Milvus

Vipshop, a major Chinese online retailer, faced significant challenges with its recommender system, which was originally powered by Elasticsearch. The system suffered from high latency, with query speeds averaging 300ms, and incurred spiking maintenance costs due to complex index management. This prompted Vipshop to seek a new solution from the vendor Zilliz to improve performance and efficiency.

Zilliz implemented its Milvus vector database to rebuild the recommendation engine. This new solution delivered a 10x faster query speed, reducing response times to under 30ms for searching millions of vectors. The measurable impact for Vipshop included a more optimized user experience with accurate recommendations, improved system scalability, and significantly reduced maintenance costs.

Vipshop