Case Study: SmartNews boosts real-time ad recommendations with Milvus from Zilliz

How SmartNews Transforms Real-Time Ad Recommendations with Milvus

SmartNews, a leading news app with high user engagement, faced technical hurdles in optimizing its real-time ad recommendations. They needed to efficiently handle high-dimensional sparse features and perform rapid unstructured data retrieval at a massive scale. Their previous solution, Faiss, was cumbersome to maintain and did not scale effectively, leading them to seek a new vendor, Zilliz, and its Milvus vector database.

Zilliz implemented Milvus, which provided the high-throughput and low-latency query capabilities SmartNews required. The solution enabled efficient similarity searches to link users with the most relevant ads, handling over 100 million daily vector updates. This led to more relevant ads being served, which increased click-through rates and boosted ad revenue for SmartNews. The team also benefited from features like collection aliasing and contributed the "hot reload" feature back to the Milvus community.

SmartNews

Dennis Zhao

AI Infrastructure Lead