Essays on the Digital Consumer: Models of Engagement, Upgrade, and Referral Behaviors

Abstract

In this dissertation, I present three essays on two empirical models of consumers using Web services. Business-to-consumer (B2C) Web services, such as Facebook, Dropbox and Pandora, have become a major part of the economy. Due to the low cost of digital distribution, these firms can provide their services for free, with the goal to attract a large customer base, while earning revenue by relying on advertisements or charging a small subset of customers for premium features. My goal is to characterize the various stages that a consumer faces when using a Web service: from adopting the service, to using it for personal and social needs, and to paying for the service. In addition, I also model the customer referral process, where the customer becomes a marketing instrument to encourage other customers to adopt. In the first essay, I explore the relationship between how customers find out about a service and how active they are when using the service. I estimate a hidden Markov model (HMM) of consumer behavior, and I characterize how the firms' social media efforts may encourage customers to be more active. In the second essay, I examine the relationships among usage, payment (upgrades), and referrals. I estimate a single agent dynamic structural model to capture these consumer decisions. Lastly, I conclude with an essay that presents the computational challenges in estimating the HMM and the dynamic structural model in a Bayesian fashion, and I also discuss how I use various estimation techniques, parallelization, and Amazon Elastic Compute Cloud to address these issues.