UA & UX in Game Development

By Xin XueUser acquisition (UA) is a key component in a game’s economic system. The most relevant question is how advertisements should be placed. Optimal advertising placement well-matches user demand and enhances user experience (UX) while maintaining profitability. Here at Happy Elements, data scientists apply parametric, time-series as well as machine learning models to predict periodic consumers’ life-time value (LTV) within each advertising campaign that measures both UX and UA efficiency. Working closely with data engineers and software engineers, data scientists design and develop an automated machine learning (AutoML) workflow to productionize and generalize LTV prediction across multiple games.

The workflow spans data extraction, transforming and loading (ETL) construction, application pattern and configuration design, feature engineering, model training, LTV prediction, and monitoring and quality assurance. Further collaborating with machine learning engineers, data scientists apply multi-armed bandit* and reinforcement learning to reallocate and optimize placement of advertisements.

* Multi-armed bandit describes the optimization problem where given limited budget and time, how we dynamically allocate them among different advertisers and targeted users to maximize user matching and our payoffs.