Data-driven customer segmentation and personalized information provision in public transit

Abstract

To ensure customer satisfaction, a transit agency must strive to understand and cater to its users' needs. The goal of this research is to develop a framework that could help the transit agency to better understand its users and their behaviors. Segmentation of the market for transit users is the first step, since it allows for the understanding of heterogeneity in their characteristics and their varying requirements, at a granular level as opposed to an aggregate one. In this study, we create a framework, which uses smart card data, to identify customer segments. The framework developed in this study includes a segmentation scheme that creates segments based on the spatial and temporal characteristics of the travel behavior of customers. Data from Hong Kong's MTR system were used to demonstrate the practical application of the developed segmentation methodology. In doing so, a thorough analysis was conducted to interpret the specifics of the identified segments. The segmentation scheme created in this study is capable of catering to meaningful applications that could serve both the agency and the users of the transit system. A few applications explored in the context of this study include the use of the customer segmentation framework for the provision of personalized information. It was demonstrated how targeted information could be provided to users who may likely be affected by a particular service disruption event. In addition, the segmentation framework was used to understand the impact of changes in the network, through a before-and-after analysis where the impact on customer travel patterns due to the provision of service on the newly opened South Island Line is adopted as a case study. Lastly, a predictive transit smart card attrition model was developed by using the features created for the purpose of segmentation. The framework for segmentation developed in this study was found to be useful for multiple applications. Furthermore, the framework is flexible and, therefore, could be generalized for use to address other applications and across other agencies.