Long-Term Sales Forecasting Using Lee-Carter And Holt-Winters Methods

Abstract

This study developed a statistical model for long-term forecasting sparkling beverage sales in the 14 provinces of Southern Thailand. Data comprised the series of monthly sales from January 2000 to December 2004 obtained from the company. We applied a classical Lee-Carter mortality forecasting approach as well as exponential smoothing Holt-Winters with additive seasonality method to log-transformed monthly sales containing season of month and branch location as factors.  The model produced excellent estimates in sales predicting for up to 24 future months of 20 branches compared with actual data in each branch during the years 2005-2006. The model also gave more accurate results than using separate forecasting method whereas it was parsimonious in the number of parameters used.