Section: New Results

Performance Evaluation of Call Centers

Participant : Pierre L'Ecuyer.

We develop research activities around the analysis and design of call centers, from a performance perspective. The effective management of call centers is a challenging task mainly because managers are consistently facing considerable uncertainty.

One aspect studied in [23] is the development of stochastic models for the daily arrival rate in a call center. Models in which the busyness factors are independent across periods, or in which a common busyness factor applies to all periods, have been studied previously. But they are not sufficiently realistic. We examine alternative models for which the busyness factors have some form of dependence across periods.

We also carry out in [14] large-scale data-based investigation of service times in a call center with many heterogeneous agents and multiple call types to investigate the validity of traditionally used standard Erlang queueing models, based on independent and identically distributed exponential random variables. Our study provides empirical support to the theoretical research that goes beyond standard modelling assumptions in service systems.

In [56], we consider a stochastic staffing problem with uncertain arrival rates. The objective is to minimize the total cost of agents under some chance constraints, defined over the randomness of the service level in a given time period. We present a method that combines simulation, mixed integer programming, and cut generation to solve this problem. In [84], we consider a particular staffing problem with probabilistic constraints in an emergency call center. We propose an algorithm to solve the problem, and validate it with a simulation model based on real data from the 911 emergency call center of Montreal, Canada.

We are also interested in predicting the waiting time of customers upon their arrival in some service system such as a call center or emergency service. In [86], we propose two new predictors that are very simple to implement and can be used in multiskill settings. They are based on the waiting times of previous customers of the same class. In our simulation experiments, these new predictors are very competitive with the optimal ones for a simple queue, and for multiskill centers they perform better than other predictors of comparable simplicity.