|
Name of Speaker |
Xiao Lei |
|
Schedule |
21 Aug 2026, 10am – 11.30am (60 min talk + 30 min Q&A) |
|
Venue |
HSS-405 |
| Link to register
(via Zoom) |
https://nus-sg.zoom.us/meeting/register/P408Y8QvSXaLCSZdXPSrtw |
|
Title |
Adaptive Design for In-App Advertising Games: A Data-Driven Methodology Validated with Field Experiments |
|
Abstract |
Rewarded ads, in which players voluntarily watch videos in exchange for in-game benefits, have become an important monetization format in both mobile gaming and digital advertising. Yet they create an operational tension: designs that increase immediate ad usage may also frustrate players and reduce future retention. We study this intertemporal trade-off in level-based puzzle games and develop a data-driven framework to maximize players’ expected lifetime rewarded-ad views through dynamic difficulty control. The problem is challenging because firms observe only limited early behavior, while player response exhibits heterogeneity in skill, ad tolerance, and churn propensity, as well as path dependence driven by difficulty reference and prior ad exposure. We formulate the firm’s decision problem as a partially observable Markov decision process with latent player types and instantiate it with a parsimonious behavioral model of win probability, ad watching, and churn. For estimation, we exploit the model structure to develop a multi-step procedure based on recursive logistic regressions. For deployment, we compute policies offline using partially observable Monte Carlo planning and compile them into compact lookup tables that satisfy sub-second latency requirements in live game systems. In collaboration with Tencent, China’s largest digital advertising network, we implement the framework in commercial games and evaluate it through two large-scale randomized field experiments. Relative to the industry baseline, the deployed policy increases long-run rewarded-ad views by 19.23% and 13.58%, while also improving engagement and retention. These results show how dynamic, data-driven experience design can serve as an implementable revenue-management lever in ad-supported digital platforms. |
|
About the Speaker |
Xiao Lei is an Assistant Professor at HKU Business School. He earned his Ph.D. in Operations Research from Columbia University. His research focuses on digital platform design, exploring how platforms can both engage and protect their users. His work has been recognized with several prestigious awards, including the Revenue Management and Pricing (RMP) Section Prize, an honorable mention for the George Dantzig Dissertation Award, and Best Paper Awards from the INFORMS Service Science and Behavioral Operations Management societies. Beyond academia, his research drives real-world impact in collaboration with industry leaders such as Tencent and Activision. |
