In-Queue Observation and Abandonment

Problem Definition: Customers arriving at a service provider are faced with a queue. On joining the queue, they get to observe the service speed while waiting in the queue. They renege if the updated wait times are too long. When and how should a revenue-maximizing service provider share service information? How does information improve or destroy consumer welfare when customers may renege? Academic/Practical Relevance: Although reneging strategies are intuitive, a rational reneging model is highly non-trivial (Hassin 2016, pp. 299). This paper models customers’ rational reneging decisions in observable queues which is void in the extant literature. Broadly, the research contributes to informationstructural issues in service operations. Methodology: Our modeling framework employs a Markovian queue with two classes of customers. Results: We fully characterize queueing dynamics in the presence of reneging customers and population characteristics. We provide analytical characterization of when a service provider should reveal the service speed information to the customers, based on reneging actions. We show that failing to account for reneging can lead to a wrong decision, and hence a significant revenue loss for the server. Interestingly, providing more information does not always benefit the customers. A calibrated provision of service speed information can realign opposing incentives and help both the firm and customers. Managerial Implications: The server’s revenue decreases and consumer welfare increases with respect to how fast customers are able to observe the service speed in-queue, but the service provider may find it beneficial to reveal the service speed information depending on the patience threshold distribution of the customers who join the queue in the first place. When the patience threshold distribution is either low or highly dispersed, it creates a stronger incentive for the service provider to reveal the service speed information. This result holds for any patience threshold distribution.