DISA SEMINAR

Predicting Time to Upgrade for Successive Product Generations: An Exponential-Decay Proportional Hazard Model

Speaker
Zhengrui Jiang, Professor, Department of Marketing and E-Business, Nanjing University
Chaired by
Dr Stanley KOK, Assistant Professor, School of Computing
skok@comp.nus.edu.sg

05 Nov 2020 Thursday, 03:30 PM to 05:00 PM

via Zoom

Abstract:
In the presence of successive product generations, most customers are repeat buyers, who may decide to purchase a future product generation before its release. As a result, after the new product generation enters the market, its sales often exhibit a declining pattern, thus rendering traditional diffusion models unsuitable for characterizing customers' time to product upgrades. In this study, we propose an Exponential-Decay proportional hazard model (Expo-Decay model) to predict customers' time to product upgrade. Compared with existing proportional hazard models, the Expo-Decay model is parsimonious and easy to interpret. In addition, our empirical test shows that the Expo-Decay model performs better than or as well as existing parametric models in prediction accuracy. Furthermore, we develop and test three extensions of the Expo-Decay model: (i) a frailty model extension that incorporates unobservable customer heterogeneity, (ii) a double Expo-Decay extension that captures influences of previous adoptions, and (iii) a time-variant extension that updates the values of covariates as time progresses. We apply the Expo-Decay model as well as the three extensions to study customers' update behaviors for a sports video game series produced by a major U.S. firm. Empirical results reveal that customers' previous adoption and usage patterns can help predict their timing to upgrade to a new product generation. In particular, we find that (i) potential switching customers who have adopted a previous product generation are more likely to upgrade; (ii) heavy players tend to upgrade earlier; (iii) specialized customers demonstrate a lower probability to upgrade. When comparing the different model extensions, we find that more complex model formulations do not lead to better prediction performance, while more accurate data does.

Bio:
Zhengrui Jiang is a professor in the Department of Marketing and E-Business, School of Business, Nanjing University. He previously was Thome Professor in Business and Professor of Information Systems in the Ivy College of Business, Iowa State University. His primary research interests include business intelligence/analytics, data quality, decision-making under uncertainty, diffusion of technological innovations, and economics of information technology. His research has appeared in leading academic journals including Information Systems Research, Management Science, MIS Quarterly, IEEE Transactions on Knowledge and Data Engineering, INFORMS Journal on Computing, and Journal of Management Information Systems. He serves, or has served, as an associate editor for Information Systems Research and MIS Quarterly, a senior editor for Production and Operations Management, and a program co-chair for the 2014 Midwest Association of Information Systems Conference, the 2015 Big XII+ MIS Research Symposium, and the 2018 Workshop on Information Technologies and Systems. He received MIS Quarterly Outstanding Associate Editor Award in 2016.