Capturing Consumer Behavioral Dynamics in Video Games: A Recurrent Marked Point Process Approach
A recent study co-authored by Eller College of Management faculty is set to be published in Marketing Science—a top journal of marketing and business research. Titled “Capturing Consumer Behavioral Dynamics in Video Games: A Recurrent Marked Point Process Approach,” the study is co-authored by Yong Liu, director of the HSLopez School of Business Analytics, professor of marketing and the HSLopez Family Endowed Chair, Junming Yin, a former assistant professor of management information systems at Eller, and Katherine Feng, assistant professor of management and marketing at the Hong Kong Polytechnic University.
The study explores the impact that content consumption—in various formats from online video games to video streaming to social media—has on modern consumers and how it has significantly influenced their purchase decisions.
Liu and his co-authors focused their research on the video game industry, which relies heavily on content consumption. As one of the most vibrant industries in the global economy, video games have a rapidly growing customer base and significant economic impact. They proposed a new framework that models and predicts content consumption (i.e., play the game) and product purchase (e.g., virtual toys, extra game features, etc.) of consumers (i.e., game players) in a video game.
The consumption of digital content products such as video games and streaming music is a continuous process in which users access the product repeatedly, and in varying durations. At the same time, users can make in-consumption purchases of additional features to enhance their experience. On top of these dynamics, digital content consumption is associated with rich consumer heterogeneity and often influenced by external factors such as the real-world backdrop of video games.
To account for these complexities, Liu and his coauthors developed a deep learning-based framework that combines an attention-based recurrent neural network (RNN) and a marked point process. The attention-based RNN flexibly incorporates both user activities and external factors to automatically learn the continuous-time representation of users’ (latent) engagement in various activities such as login, content consumption and in-consumption purchase. Based on these representations, the marked point process jointly characterizes the occurrence time, consumption rate and purchase count for each user activity.
Calibrating and testing the framework in the context of a major sports video game, they found this new model outperforms a host of benchmarks in predicting players’ game-play and purchase behaviors. With model parameters, the co-authors also estimate each player’s base propensities of playing and purchasing and show the significantly different behavioral patterns across player segments. Econometric analysis results further demonstrate that both past in-game activities and external sports matches are positively associated with increases in players’ login frequency, game-play duration and purchase count, and such effects vary across players with different game-play and purchase propensities.
Their results suggest that players’ past game-play duration and purchase count have an overall positive effect on subsequent activities (i.e., login, game-play and purchase). However, not all behavior outcomes are affected to the same degree. Players show a consistent behavioral pattern across time, meaning that the same behavior will carry over in the future, such as continuing playing or purchasing in the game. More importantly, the results show that in-game purchase is a different motivation from pure playing behavior. The purchase in the video game is an in-consumption behavior, for the purpose to enhance playing performance.
Based on their research, Liu and his co-authors suggest that video game companies should target “gamer” and “casual” segments more, rather than “hardcore” or “buyer” segments who are the players who have purchased enhancement packs in the past, are more likely to log in and play the game in the future, as they need to spend what they bought.
They suggest the video game company could design effective marketing promotions to induce the “gamer” and “casual” segments to purchase in-game packs, such as offering larger product discounts to these two groups.
The study includes future directions that could include further investigation of this topic beyond the context of their model.