Evaluating the fit of the Erlang A model in high traffic call centers

We consider the Erlang A model, a queuing model often applied to analyze call center performance. While not a new model, Erlang A is becoming a popular alternative to the widely used Erlang C model. In this paper we analyze the accuracy of Erlang A predictions in high traffic environments, a situation where the Erlang C model is not applicable. Our findings indicate that in this high traffic region the Erlang A model is subject to a moderate to high level of error that has a strong pessimistic bias; that is the system tends to perform better than predicted. This is in sharp contrast to lower volume scenarios where the model tends to be optimistically biased. We find that in addition to utilization, the model is most sensitive to arrival rate uncertainty and balking.