Abstract
Service companies often adopt on‐demand scheduling to balance labor costs and fluctuating market demand. However, research shows that such practices can reduce worker productivity and retention. In this study, we examine how on‐demand scheduling affects two critical outcomes: worker absenteeism and patient dissatisfaction. We extend the conceptualization of undesirable scheduling by introducing schedule discontinuity —the presence of unpaid interruptions within a worker's daily schedule—alongside the more commonly studied schedule inconsistency , or variability in work hours across weeks. Using data on 1.2 million home care visits in a Canadian healthcare provider, we find that both schedule inconsistency and discontinuity significantly increase absenteeism and patient dissatisfaction. Specifically, moving from the 25th to 75th percentile in discontinuity (inconsistency) raises absenteeism by 20.00% (19.29%), and customer complaints by 27.33% (40.27%). To assess the practical implications for employers, we formulate and solve a schedule optimization problem that minimizes schedule discontinuity (or inconsistency), while satisfying demand and supply constraints. Applying a machine learning predictive model to these optimized schedules, we estimate reductions in the probability of absenteeism by 9.5% (8.2%) and in the probability of patient complaints by 7.7% (2.3%), demonstrating that modest scheduling adjustments can substantially improve worker and service outcomes.