Service-Level Variability of Inbound Call Centers

In practice, call center service levels are reported over periods of finite length that are usually no longer than 24 hours. In such small periods the service level has a large variability. It is therefore not sufficient to base staffing decisions only on the expected service level. In this paper we consider the classical M/M/s queueing model that is often used in call centers. We develop accurate approximations for the service-level distribution based on extensive simulations. This distribution is used for a service-level variability-controlled staffing approach to circumvent the shortcomings of the traditional staffing based on the expected service level.