Selecting the appropriate input data set when configuring a permanent workforce

This paper explores the issue of choosing the best data to use when running a scheduling model to select a permanent workforce for a service facility. Because demand is assumed to vary over the day and throughout the week, the choice of the input data is crucial. If a week of low volume is selected, the solution might call for an insufficient number of workers; if a week of high volume is chosen, excessive idle time might be the result. Staff scheduling at mail processing and distribution centers (P&DCs) in the United States provides the backdrop. In operating these facilities, a critical objective is to manage overtime, part-time workers, and temporaries so that when volumes are high, additional costs are kept to a minimum, and when volumes are low the permanent workforce is almost never idle. In quantitative terms, this means selecting the size and composition of the workforce so that over the year, no more than a total number of overtime hours, part-time hours and temporary worker hours are used when demand exceeds some baseline and absenteeism is taken into account. To solve the problem, an engineering approach is proposed in which estimates of productivity are made based on a single run of the optimization model and the final data set is chosen to satisfy a small error tolerance. The full methodology is illustrated with data provided by the Dallas P&DC.