A Comparative Evaluation of Forecast Monitoring Systems in Service Organizations

Demand forecasts are major inputs to workforce scheduling and material planning in many service organizations, and the effectiveness of such planning activities hinges upon the accuracy of the forecasts. Since forecasts are rarely precise in reality, managers need to monitor forecast errors when they implement the labor and material plans. This paper aims to identify and evaluate automatic detector of forecast bias to help managers. This paper identified and evaluated five error detection techniques using both actual data from a call center, and simulated data. All five techniques detected a considerable demand shift in a timely manner, and appeared very robust across diverse demand environments. In particular, Threshold curve and Wineglass chart turned out to be the quickest and most powerful of the five methods. In addition, the patterns of within day demand arrival and their stability throughout the day significantly influenced the performance of the detection techniques.