Speaker "Sampsa Jaatinen" Details Back



Deep reinforcement learning for free to play mobile game monetization optimization.


Free to play games have become the largest and fastest growing mobile game market. In free to play model the game itself is free for download and gameplay, but the developer accrue revenues by selling virtual goods, extra play time or other content as add on or by advertising. Monetization in these games relies on two main avenues, one being game players paying for content and second being advertisers paying for time with the game player. Mobile advertising typically consist of static image like advertisement, video advertisement or interactive assets where viewer can e.g. engage by playing mini-version of the advertised game or other application. Unity’s Monetization serve personalized content to hundreds of millions of mobile devices on monthly basis. In this talk we introduce a Q-learning based deep reinforcement learning approach to optimize for mixed Advertisement and IAP promotion content delivery for optimal strategy. We show, how one-step optimization policy can be improved towards to better engagement and monetization. We display results from real games using Unity’s services for monetization.


Dr. Jaatinen has worked with big data, predictive modelling and machine learning since 2008, with a focus on mobile technology and applications usage. He has established machine learning as part of the Unity Technologies Monetization services and leads teams optimizing the advertising and in-app purchase solutions.