Forecasting Teller Window Demand with Exponential Smoothing

Abstract

This paper demonstrates the use of exponential smoothing to forecast the total daily demand for teller window services in a bank. It describes (1) the procedures that were used to detect the underlying causes of systematic variation in teller window demand and (2) the use of an exponential smoothing model with seasonal factors that are not time dependent. Experimental results are presented which illustrate the improvement in forecasting performance obtained by using the nontime-dependent seasonal factors. The resulting forecasts provide a means of improving teller window staffing decisions.