Predicting queueing delays for multiclass call centers

This paper deals with real-time operation management in call centers. We investigate the prediction of the waiting time for each customer upon arrival. We model our call center by a multiclass multiserver service system. The system consists of a pool of full-flexible agents, and two classes of impatient customers. First, we propose estimators for the state-dependent waiting time distributions for new arrivals by incorporating the change that may occur on their waiting experience. Next, we investigate the way to inform customers about their anticipated delays. We discuss different ways to communicate this information. In particular, we give attention to analyzing the announcement of the waiting time by steps.