10:30 AM - 10:55 AM
Forecasting Call Center Arrivals: Bivariate and Mixed Models
We consider different statistical models for the call arrival process in telephone call centers. We evaluate the forecasting accuracy of those models by describing results from an empirical study analyzing real-life call center data. Our study shows the importance of accounting for alternative correlation structures in the data.
10:55 AM - 11:20 AM
Forecasting and Staffing Call Centers with Multiple Uncertain Demand Streams
We consider the entire process of forecasting and staffing call centers with multiple uncertain demand streams. We first develop methods to simultaneously forecast multiple stream arrivals by incorporating the dependence between different streams. Given the forecasts, we then implement a chance-constraint staffing algorithm to generate staffing vectors and further assess the operational effects of incorporating dependence among different streams. Experiments using a real call center data set show that the multiple-stream approach provides more accurate forecasts, and the corresponding policy meets the quality of service objective better, compared with the single-stream approach. Simulation studies indicate that the forecasting and operating benefits of multiple-stream approach vary by type and strength of the dependence among streams.
11:20 AM - 11:45 AM
Weight-Based Routing for Multi-Skill Call Centers using Call Waiting Times and Agent Idle Times
We propose a new routing policy for multi-skill call centers based on weights, call waiting times and agent idle times. The routing is controlled by additive and multiplicative weight parameters that are optimized by simulation-based metaheuristics. The quality of service constraints are formulated as penalty cost functions. Numerical examples show that our routing policy often performs better than those used in practice.
11:45 AM - 12:10 PM
Sensitivity of the Pooling Decision to the Dependence Across Call Types in the Hydro-Québec Large Call Center
We specified copula-based multivariate models to capture the dependence structure across call types arrival processes. We showed that ignoring or mismodeling the dependence structure can lead to erroneous simulation based estimation of this system performance. We provided managerial insights for choosing which call types to merge on the same pool.