Scheduling Agents Using Forecast Call Arrivals at Hydro-Québec's Call Centers

The call center managers at Hydro-Québec (HQ) need to deliver both low operating costs and high service quality. Their task is especially difficult because they need to handle a large workforce (more than 500 employees) while satisfying an incoming demand that is typically both time-varying and uncertain. The current techniques for determining the schedule of each employee according to the forecast call volumes at HQ are often unreliable, and there is a need for more accurate methods. In this paper, we address the concerns of the call center managers at HQ by modeling the problem of multi-activity shift scheduling. This model has been implemented and tested using real-life call center data provided by HQ. The main contribution of this paper is to demonstrate that a constraint programming (CP) model with regular language encoding can solve large problems in an industrial context. Furthermore, we show that our CP-based formulation has considerably better performance than a well-known commercial software package.