Key determinants and heterogeneous frailties in passenger loyalty toward customized buses: An empirical investigation of the subscription termination hazard of users
Introduction
An emerging demand-driven, user-oriented transit service, known as a customized bus (CB) service, holds the promise of alleviating urban traffic congestion, especially under the extremely tight road capacity constraints during peak hour. In contrast to conventional public transportation (PT) services, CBs aggregate users with similar origins, destinations, and departure/arrival times (Liu and Ceder, 2015). The aim of a CB service is to provide point-to-point transit services (with one person per seat) without any transfers between the appointed boarding and deboarding stations. Thus, user requests are assigned as a top priority. The CB service has gained popularity in many large and medium-sized cities in China since it was first launched in Qingdao (August 2013), similar to other on-demand ride services (Ke et al., 2017).
A profound understanding of the subscription behavior is vital not only for service improvement but also for maximizing user retention. Due to the high operation costs, CB routes with high seat occupancy rates, for example, greater than 80%, have been launched. CB routes with relatively low seat occupancy rates are partially adjusted, combined, or may be canceled if the occupancy rate in operation is less than 50% for several months. The survival of a CB route heavily depends on the stability of the subscription quantity and the duration for each user, which represent the level of user loyalty. In reality, even though the service providers have made concessions on the occupancy rate and developed strategies such as combining similar routes to retain passengers, there have been drastic decreases in the number of CB service users after its prevalence in many cities (Yan, 2014, Chu, 2015, Ma, 2015), such as Guangzhou, Hangzhou, Nanjing, Xiamen, and Harbin. Many launched routes have been canceled, and many users have unsubscribed from the service and returned to their previous modes of transportation, leading to a substantial waste of resources and the disappointment of many loyal patrons. The drastic decrease in the number of CB users has been mainly attributed to the following reasons: 1) passengers gradually stopped using the service after seeking novelty with regard to this new concept of personalized transit service or just finishing the follow-the-leader behavior; 2) the absence of demand forecasting techniques and flexible route designing methods that could satisfy the rapid increase in new users, while the launched routes could not attract enough passengers to maintain operation without discount coupons; and 3) the prereservation of the CB service and compliance with the predetermined schedule, which is not as flexible as using a conventional high-frequency PT system.
Essentially, these decreases show the vicious transformation of users’ loyalty toward the service that they have subscribed to. Ensuring the survival of the launched routes is the only win-win solution for both operators and loyal users. Feedback from the viewpoint of user subscription behavior is a good indicator of user satisfaction toward the supplied service and will provide practical references for policy-making and market orientation. Factors leading to the loss of loyalty can be targeted with intensive strategies to maximize user retention. Therefore, ascertaining the key determinants of passenger loyalty, as well as modeling and forecasting the subscription duration, are urgent and critical events in the development of the CB system.
There is a paucity of studies that focus on investigating passenger loyalty directly by using their actual purchase records (Trépanier et al., 2012). Additionally, there are even fewer studies on user loyalty toward the CB transit service. This work aims to fill this gap by investigating CB user loyalty through a survival analysis based on their long-term subscription behaviors. The survival models, or so-called duration models, are employed to model user loyalty with factors drawn from historical subscription behaviors, travel attributes, and individual characteristics. The use of the survival model enables the analysis of factors that mainly drive users’ loyalty and facilitates the investigation of the mechanism of subscription behaviors. From the viewpoint of system operation, analyzing passenger survival time is an important benefit because it can help measure user loyalty at different stages and respond to strategy changes for further improvement of the service.
The rationale of the CB service is illustrated in Fig. 1. The CB aggregates users with similar origins, destinations, and departure/arrival times, which can be clustered by demarcating the demand distribution into OD bundles. Each OD bundle is composed of an origin zone, a destination zone, and a passing-by corridor. The ladder-shaped origin/destination demarcations are beneficial for shortening the detour distance by picking up and dropping off passengers. The boarding and alighting stations are designed to be located close to the users’ residences and workplaces. Moreover, there are no stations located in the span of the passing-by corridor, thereby making the route flexible and ensuring that the drivers follow the vehicle navigation system to minimize the travel time by following the shortest-time path. Users can subscribe to CB services by using easy-to-access websites and smartphone applications. The CB service has rapidly become popular in China due to its simplicity, convenience, comfort, and customized features.
The remainder of this paper is organized as follows. Section 2 reviews key studies relevant to the investigation and modeling of passenger loyalty in transit systems. Detailed descriptions of the study area and the data collection of the CB system are given in Section 3. The methodology employed in this study and the models built are described in detail in Section 4. Finally, the results and discussions, as well as the conclusions drawn based on the key findings of this study, are presented in Section 5 and Section 6, respectively.
Section snippets
Literature review
Customer loyalty toward a specific product or service is defined as the repetitive purchase behavior that reflects the decision to continue buying the same product or service (Jacoby and Chestnut, 1978). Customer loyalty can be classified into behavioral loyalty and attitudinal loyalty (Webb, 2010, Odin et al., 2001). Behavioral loyalty is a person’s repetitive selection of a certain brand over the competition (Odin et al., 2001), whereas attitudinal loyalty is associated with the psychological
Research area and data collection
The study area chosen for this work is Dalian, a coastal city in Northeast China. There are different PT modes available in Dalian, such as subways, light rail transit, and regular buses. The rail transit system in Dalian is not well developed. There are two subway lines that primarily connect the functional areas of the city center, whereas the other two light railway lines connect emerging city outskirt centers to the main city areas.
In March 2014, the first CB route was launched in several
Loyalty of CB users
A survival analysis uses a collection of statistical procedures for data analysis, for which the outcome variable of interest is the time until an event occurs (Kleinbaum and Klein, 2010). It is popularly used in many areas such as medicine, biology, engineering, and economics. In this study, the event is the termination of the CB service subscription, which indicates the loss of loyalty from a specific passenger. When a user has stopped subscribing for a long time, usually longer than his/her
Results and discussions
Fig. 4 is a K–M survival curve showing the cumulative proportion of the CB users retained over time. The survival probability initially decreases faster and levels off after approximately 22 months. The results validate our assumption that the first riding experiences are more critical for user loyalty than the repeated ones.
The models built above were demonstrated significantly similar parameter estimates. Thus, only the parameter estimates of the shared frailty model are listed in Table 3,
Conclusions
To fill the research gap in user loyalty for the CB service, survival models were employed to determine user loyalty and ascertain key determinants and heterogeneous frailties based on long-term subscription records as well as OD information on CB users in Dalian, China. The use of the survival models enabled analyzing factors that mainly drive user loyalty and facilitated investigating the mechanism of continuous subscription behaviors. The key findings are concluded as follows:
- 1.
The factors
CRediT authorship contribution statement
Jiangbo Wang: Conceptualization, Methodology, Software, Visualization, Investigation, Writing - original draft, Writing - review & editing. Toshiyuki Yamamoto: Conceptualization, Methodology, Supervision. Kai Liu: Conceptualization, Data curation, Investigation, Writing - original draft, Writing - review & editing, Supervision.
Acknowledgments
The author thanks the anonymous reviewers and the editor for their insightful comments that improved the paper substantially. The first author graciously acknowledges the China Scholarship Council for its financial support. This work was carried out by the joint research program of the Institute of Materials and Systems for Sustainability, Nagoya University. The last author would like to acknowledge the support funding from the National Natural Science Foundation of China (Grant Nos. 51378091
References (51)
- et al.
A model to assess public transport demand stability
Transp. Res. Part A Policy Pract.
(2011) - et al.
An ontology-based framework to support performance monitoring in public transport systems
Transp. Res. Part C Emerg. Technol.
(2017) Development of an accident duration prediction model on the Korean Freeway Systems
Accid. Anal. Prev.
(2010)- et al.
Identifying potential market niches for Demand Responsive Transport
Res. Transp. Bus. Manag.
(2012) - et al.
Anonymizing trajectory data for passenger flow analysis
Transp. Res. Part C Emerg. Technol.
(2014) - et al.
Time-varying effects in the analysis of customer loyalty: A case study in insurance
Expert Syst. Appl.
(2012) - et al.
Factors influencing customer loyalty to educational tour buses and measurement invariance across urban and rural zones
Transp. A Transp. Sci.
(2015) - et al.
Nonparametric identification of daily activity durations using kernel density estimators
Transp. Res. Part B Methodol.
(2002) - et al.
Behavioral intentions of public transit passengers-The roles of service quality, perceived value, satisfaction and involvement
Transp. Policy
(2011) - et al.
Impact of road gradient on energy consumption of electric vehicles
Transp. Res. Part D Transp. Environ.
(2017)