Forecasting Weekly Demand Of Coffee – Based Categories For A Coffee Chain In Ho Chi Minh City: A Case Study

Abstract

This study proposes a procedure for selecting forecasting models to support a coffee store chain in choosing suitable techniques for their business. It compares traditional and modern statistical methods for predicting the 4 – week heading demand of two coffee – based categories, Espresso and Vietnamese Coffee. The quantitative techniques employed include SES, ARIMA, LS, MLR, DT, and RF. DT and SES are identified as the optimal methods for Espresso and Vietnamese Coffee, respectively, based on error measures such as MAE, sMAPE, RelMAE, MdAE, sMdAPE, and RelMdAE, with the mean method as the benchmark. Additionally, the study suggests the optimal weekly order quantity for coffee packages using the newsvendor model to maximize profit.