The Effects of Information Granularity on Abandonment and Congestion in Non-Stationary Priority Queues

Motivated by empirical studies on the impact of information on abandonment in service systems, we study novel queueing models with abandonment rates that depend on the system state and the granularity of customers’ state information.We consider a two-class multi-server priority queue with time-varying arrivals under three levels of information granularity: no, partial, and full information. For each level, we establish the fluid approximation of the queue length process, and equilibrium existence, uniqueness, and stability for the fluid model. We then compare the equilibrium performance under the three information levels and generate insights on (1) the effects of information granularity on the queue length and abandonment rate, and (2) how these effects depend on time-varying arrivals and priority service: In the single-class case, no information minimizes the queue-length, whereas full information minimizes the abandonment rate under non-stationary arrivals; hence, information design involves a trade-off between these metrics. Priority service yields further trade-offs between the priority classes: no information minimizes the High-Priority (HP) queue length whereas partial or full information minimizes the Low-Priority (LP) queue length; further, under non-stationary arrivals, full information minimizes the HP abandonment but no information typically minimizes the LP abandonment.These findings have important information design implications: (i) It is important to properly weigh the costs and benefits of queueing and abandonment, and to evaluate the benefits of time- and/or state-dependent information regimes. (ii) With priority service, information design should consider class-specific service-level targets and/or financial metrics, and evaluate the benefits of class-dependent information regimes