Transformed Business Model and Cost Structure slide image

Transformed Business Model and Cost Structure

Cutting-edge search and advertising technology Y 8 Experimentation driven Hundreds of product experiments running simultaneously, with thousands of experiences tested per year Broad-based Local Ad Platform High velocity Median developer lead time per feature improved by 80%³ Sophisticated ML & deep learning power advanced functionality • Yelp's unsurpassed consumer review data yields personalized search and recommendations Deep understanding of our data allows prediction of user queries and optimization of our SEO strategy Our state-of-the-art search platform outperforms the industry-standard solution by 30-50% while costing as much as 40% less to run, allowing better ad targeting and search matching To conduct an average of 20M auctions per day, we predict demand for 16M ad categories and time intervals to set bid pricing and pacing, then optimize bids 48x per day per advertiser with the goal of delivering more value Novel "small-block" deep learning infrastructure yields 100x training speed increase² for better click-through rate prediction Proprietary service orchestration framework allows dynamic and scalable AWS utilization with class-leading cost efficiency 1 Compared to Yelp's previous Elasticsearch-based ranking platform. Cost savings consist of reduced infrastructure costs. 2 Compared to the training speed of traditional deep learning frameworks on this problem. 3 Improvement measured in 2020 and sustained into 2021. Includes U.S. employees only 16
View entire presentation