Slow First, Fast Later: Temporal Speed‐Up in Service Episodes of Finite Duration

Many service environments such as outpatient departments comprise repeated episodes of finite duration wherein customers arriving during an episode must be served before the end of that episode. Scheduling resources and customer arrivals in such settings involves trading off customer wait against resources' time and effort. We hypothesize that this trade-off leads to a "slow first, fast later" pattern of system speed, which allows build-up of inventory earlier for more efficient utilization of faster system speed later. As a natural corollary, we also hypothesize that greater anticipated workload, which causes faster inventory build-up, leads to a larger increase in system speed earlier in the service episode than later. We empirically validate these hypotheses using operational data from a high volume tertiary care outpatient department. Our estimation results suggest that due to hypothesized system speed pattern, patients arriving around the middle (later) part of the episode experience 20% (63%) shorter length of stay compared to those arriving at the beginning. Further, additional anticipated workload of 1 patient per hour leads to 14%, 4%, and negligible reduction in the average length of stay during the earlier, middle, and later part of the service episode, respectively. Counterfactual simulations based on our model estimates show that "slow first, fast later" system speed pattern yields faster completion of work as compared to an equivalent constant system speed pattern.