Análise comparativa de métodos de previsão de demanda: estudo aplicado em uma empresa varejista de Mato Grosso do Sul

Abstract

This study investigates the stock management issue at a retail company specializing in commercial refrigeration and food service equipment located in Mato Grosso do Sul, which bases its purchasing decisions predominantly on managerial intuition, resulting in an unbalanced inventory and financial loss. The general objective was to perform a comparative analysis of simple demand forecasting methods to identify the most suitable models for the company’s product demand behavior. The research is characterized as applied, with a quantitative approach and exploratory objective. The product portfolio was categorized into three distinct profiles based on added value and sales volume. Using the ABC Curve method with sales quantity and revenue parameters, one representative product was selected from each category. Monthly demand historical data were collected from July 2022 to September 2025, totalling 39 records per product. Three forecasting methods were applied: Simple Moving Average, Weighted Moving Average and Simple Exponential Smoothing. Model performance was evaluated using Tracking Signal, MAPE and MAD metrics. Results demonstrated that no universally superior method exists for all products. For product P1, the Weighted Moving Average with a 3-month interval showed the best performance. Product P2, despite achieving the best result with 3-month Simple Moving Average, presented a MAPE of 290.76%, indicating high volatility and inadequacy of simple methods for its forecasting. For product P3, the Simple Moving Average with a 3-month interval was most suitable. The study concludes that the choice of ideal method depends directly on each item’s historical demand profile, and that products with high variability require more sophisticated techniques than traditional time series methods.