Assessment of the Sales Forecast Technique Double-Weighted Moving Average vs Other Widely Used Forecasting Techniques

Abstract

In order to improve the operations planning of two companies, whose main business is to be chemical products suppliers in Mexico, it was made the sales forecast of a fourth year of operations, using the monthly sales data information of the three previous years. The objective of the chemical suppliers forecast was to be in a better position to satisfy the multiple and varied needs of their clients, which demand different quantities of products and have different consumption patterns. The sales forecast was made by the next six techniques: Simple Moving Average (SMA), Weighted Moving Average (WMA), Trend Projection (TP), Exponential Smoothing (ES), Simple Linear Regression (SLR), and the recently proposed (Castillo, et al. 2016) technique called: Double-Weighted Moving Average (DWMA). The three years monthly sales data of 61 products, handled by the two companies, were processed in order to obtain the monthly forecast of the fourth year. After the fourth year, the forecasted data were compared with real monthly sales data. The analysis was made by the determination of the Symmetric Mean Absolute Percentage Error (SMAPE), which gave the next results: In the case of company 1, the average errors for the five reference techniques (SMA, WMA, TP, ES and SLR) was in the range 0.235 – 0.351] vs 0.249 for the DWMA. For company 2, the average error, of the same five reference techniques was in the range [0.292 – 0.467] vs 0.282 for the DWMA. WMA was the second technique in giving the least forecasting errors. In both companies, DWMA was the forecasting technique with one of the lowest average error and the lowest error in most of the products.