How Is Mobile User Behavior Different? — A Hidden Markov Model of Mobile Application Usage Dynamics

Abstract

Mobile application usage is becoming an essential activity in many people’s daily lives. Compared with PC internet, mobile internet usage is ubiquitous, temporally fragmented, and more context-dependent. Thus far research is limited on the mechanism of mobile app usage and the effects of context, even as usage continues to grow. In this paper, we aim to develop a framework to capture the underlying mechanism of mobile app usage, taking into account its unique time-fragmented feature and the possible impacts of contextual factors. To this end, we propose a hidden Markov model (HMM) and calibrate it using a consumer panel that contains real-time app usage information. We find three hidden states driving mobile app usage: utilitarian, social, and hedonic. Consumers have multiple intentions in the utilitarian state but are relatively single-minded in either the social or hedonic state. The state dynamic is volatile: Chains of continuous utilitarian states and hedonic states intercommunicate frequently with densely intermittent social states. Such a volatile state dynamic provides an antecedent for the fragmented-time mobile usage phenomena. Contextual factors—in particular, location and time of day—influence the state dynamic, e.g., its volatility varies across different locations and times of day. In sum, our analysis depicts the following picture: In the mobile internet era, consumers take advantage of relatively long free-time windows for functional and entertainment activities, while exploiting short micro-moments for social activity. Furthermore, with the help of mobile internet technology, consumers seem to: (1) utilize more micro-moments outside the office or in the morning as a complement to work, and (2) reserve more micro-moments in the evening for relaxation and entertainment, all at the expense of social activities.