Transforming Ad Recommendations: SmartNews's Journey with Milvus
In the fast-paced world of news apps, SmartNews has emerged as a dominant force, capturing user engagement with unparalleled success since its establishment in 2012. Headquartered in Tokyo, Japan, and making significant strides in the U.S., SmartNews faced a unique challenge – optimizing ad recommendations for its highly engaged user base. The journey from challenge to triumph involved overcoming technical hurdles, exploring innovative solutions, and ultimately finding success with Milvus.
SmartNews required real-time personalized ad recommendations for its soaring user engagement.
SmartNews's user engagement was soaring, surpassing industry giants like Apple News and Google News in average monthly user engagement as of July 2021. However, this success brought forth a new challenge — delivering timely and relevant ads to a discerning audience. Shu, an AI infrastructure engineer at SmartNews, detailed the technical hurdles: high-dimensional sparse features and the need for rapid unstructured data retrieval in their large-scale machine learning systems.
The existing solution, FAISS, proved cumbersome and inefficient, notably as the platform expanded. Maintaining high standards of user engagement required a solution that could handle the complexities of their data and provide real-time ad recommendations.
Milvus makes a vast difference in optimizing SmartNews’s ad recommender
SmartNews turned to Milvus after a thorough two-month research phase seeking a solution that could handle high-throughput and low-latency queries – critical for a real-time ad recommendation system.
Milvus's vector similarity search capabilities became instrumental in optimizing SmartNews's dynamic ad vector recall. During the POC phase, Milvus met SmartNews’ strict requirements of P99 latency < 10ms, which is crucial for our real-time ad recommendation system. Shu, one of SmartNews’s engineers, highlighted that Milvus allowed the team to map incomparable data types into vectors, enabling efficient similarity search to link users with the most relevant ads. This capability proved essential as SmartNews processed millions of vectors in real time to deliver personalized ad recommendations.
Beyond ad recommendations, Milvus showcased its versatility in other machine-learning applications. SmartNews team also used Milvus for real-time data and index updates, which was vital for keeping the ad recommendation models current.
Recognizing the value of community-driven development, SmartNews actively contributed features like 'hot reload' to the Milvus community, enhancing both internal operations and the broader developer community.
Adopting Milvus: from technical upgrade to tangible success
The adoption of Milvus wasn't just a technical upgrade; it translated into actual business results for SmartNews. Yusup, another engineer at SmartNews, emphasized that Milvus's collection aliasing feature streamlined the update process, addressing the cumbersome mechanism for updating vectors.
The optimized ad recommendation system powered by Milvus led to more relevant ads, increasing click-through rates and ultimately driving up ad revenue. In addition, real-time data and index updates empowered SmartNews to adapt swiftly to changing user behavior, ensuring consistently relevant recommendations.
SmartNews has upgraded its Milvus to 2.2.4 and is now gearing up for its next technological evolution phase. Dennis Zhao shared, "The new version promises even better performance and scalability, and we're excited about leveraging the new features to build even more real-time and reliable systems."
Summary
In news apps, SmartNews exemplifies how strategic technological choices, like embracing Milvus, can not only overcome challenges but also propel a platform to new heights of user satisfaction and business success. The synergy between innovation and collaboration has positioned SmartNews as a trailblazer, setting the stage for continued excellence in real-time ad recommendations.
Fendy FengFendy Feng is the Technical Marketing Writer at Zilliz. She has extensive experience developing and enhancing the impact of open-source projects in various global markets by producing high-quality, tailored content. Before joining Zilliz, Fendy worked as a Content Strategist at PingCAP, a fast-growing E-Series startup renowned for its open-source distributed SQL database.
- SmartNews required real-time personalized ad recommendations for its soaring user engagement.
- Milvus makes a vast difference in optimizing SmartNews’s ad recommender
- Adopting Milvus: from technical upgrade to tangible success
- Summary
Content
Start Free, Scale Easily
Try the fully-managed vector database built for your GenAI applications.
Try Zilliz Cloud for Free