Prescriptive Call Center Staffing

Abstract

This paper addresses the call center staffing problem. We present a novel prescriptive staffing approach that minimizes the human labor cost and the cost for calls that were abandoned due to excessive waiting times.

Our approach is novel in that it determines a prescriptive model based on the functional relationship between observable features such as call volumes in previous staffing segments, school holidays or other events and the optimal staffing decision. In order to abstain from strong assumptions about underlying data distributions, we learn the model from historical data by combining the staffing cost optimization problem with a machine learning algorithm.

We analyze the performance of our approach on two real-world data sets and compare it to a state-of-the-art benchmark.

Provided with the same information as the benchmark, our approach dominates on both data sets, resulting in a cost improvement of up to 8 percentage points and shows even greater cost improvements when provided with additional features.

We can explain the cost advantage of our approach in part with its ability to consider non-random intra-slot patterns in the call arrival such as a trend.

Keywords: Capacity Management, Machine Learning, Call Center, Fluid model,