User Growth Algorithms and Engineering Practices at Huya Live Streaming

Introduction: Huya, a leading Chinese live streaming platform, focuses on user growth through advertising‑driven acquisition and retention.

Background: Describes the full user growth funnel (acquisition, activation, retention, revenue) and the ad placement workflow, including target audience definition, bidding, click‑through, activation, and quality monitoring.

Algorithm practice: Covers four key components—crowd targeting, material selection, bidding strategy, and user acquisition strategy. Details three stages of crowd targeting (offline whitelist, whitelist + blacklist, real‑time bidding (RTA) control), uplift modeling to identify marketing‑sensitive users, virtual callbacks to improve conversion reporting, and the use of XGB‑plus‑Logistic‑Regression model fusion.

Engineering practice: Describes material automation and selection, automatic cropping, automated ad configuration and multi‑platform deployment, RTA flow filtering with low‑latency (<10 ms) response, local storage of black‑ and white‑lists using Bloom filters and FlatBuffers, incremental and full‑load updates, dynamic operator pipelines for multiple strategies, and click‑through attribution via virtual callbacks.

Results: Virtual‑callback and uplift strategies significantly improve DAU, retention, and cost efficiency; intelligent bidding reduces waste and raises user quality; the system has been deployed across Huya’s products and serves as a foundation for future automated, data‑driven growth solutions.