Slow First, Fast Later: Empirical Evidence of Speed-up in Service Episodes of Finite Duration

Abstract

There is increasing evidence that workers in many service settings regulate their work speed in response to operating conditions. However, in service episodes of finite duration characterized by time-varying and non-stationary dynamics (e.g., outpatient departments), operating variables that can affect work speed have not been rigorously studied. Understanding the impact of these drivers on work speed can help managers improve resource and appointment scheduling. We employ the trade-off faced by workers between cost of providing service and cost of customer wait in such environments to identify two previously unexplored drivers of work speed - time within the episode and anticipated remaining workload. We empirically test our predictions using data from a high volume, tertiary care outpatient department. We find that workers start a typical service episode at slower work speed and progressively speed-up toward the end of the service episode. As a result, incoming patients experience 23% shorter (50 minutes smaller) average length of stay at the middle, and 60% shorter (130 minutes smaller) toward the end of the service episode. Further, consistent with our conceptual framework, we find that the temporal pattern of speed-up is influenced by the anticipated workload during the service episode. Specifically, a higher anticipated workload leads to a larger increase in work speed earlier in the service episode than later. An increase in anticipated workload by 1 patient per hour leads to 9% (20 minutes) reduction in average length of stay at the beginning, only 4% (8 minutes) reduction during the middle and negligible change toward the end of the service episode. We also identify the role of reduction in discretionary task (ordering diagnostic tests) as one mechanism of achieving this speed-up, which itself is influenced by time within the service episode and anticipated remaining workload. These findings provide managers with several levers to improve efficiency and quality of service provision, e.g. overall shorter service episodes, increasing the staffing level earlier in the service episode and decreasing it later, and scheduling simpler patients earlier in the service episode.

Keywords: Non-stationary and time-varying dynamics, Discretionary Tasks, Finite Duration