Exponential Smoothing Methods for Demand Forecasting in a Fast-Moving Consumer Goods Company in Lagos

Accurate demand forecasting is a cornerstone of efficient supply chain management in the fast-moving consumer goods sector, where stockouts and overstock situations both impose significant costs on manufacturers and distributors. This study evaluates and compares simple exponential smoothing, Holt's two-parameter method, and Holt-Winters' triple exponential smoothing for weekly demand forecasting of five product categories in a consumer goods company operating from Lagos, South West Nigeria. Historical weekly sales data covering January 2021 to December 2023 are obtained from the company's enterprise resource planning system. The data exhibit varying degrees of trend and seasonality across product categories, motivating the evaluation of all three exponential smoothing variants. Smoothing parameters are optimised by minimising sum of squared forecast errors over a 104-week training sample, with 52-week holdout validation. Forecast accuracy is evaluated using mean absolute error, mean absolute percentage error, and tracking signal. Results indicate that Holt-Winters' method achieves the lowest mean absolute percentage error across four of five product categories, with values ranging from 7.1 to 11.4 percent. For the single category exhibiting no discernible seasonality, Holt's two-parameter method is optimal. The study recommends integrating the selected methods into the company's inventory management system with automatic parameter re-estimation every quarter. Keywords: exponential smoothing, demand forecasting, fast-moving consumer goods, Lagos, Holt-Winters.