Efeitos da previsão de demanda sobre os custos de estoque em uma empresa do varejo de confecções e calçados: otimização via simulação de Monte Carlo.

Abstract

The retail market in Brazil holds significant relevance within the national economy, being the sector with the largest number of active companies in the country. In retail, the ability to forecast demand with accuracy and consistency represents an important management practice, as it enables the development of more assertive strategies, such as proper inventory control. However, demand forecasting can be considered a challenging activity, as it involves multiple variables and statistical analyses, especially when the goal is to evaluate data with fluctuations. The purpose of this research is to assess the impact of product family demand forecasting on the inventory costs of the company São Luiz Moda Griffe. This is a field study, with data collected from the company’s records database. The methodological approach is quantitative, since the analyses were developed using mathematical models, statistical tools, simulation, and optimization. Regarding its objectives, the research is classified as exploratory-normative, as it sought to define the most suitable forecasting models through the calibration of optimal parameters, in addition to testing alternatives to avoid stockouts or excessive inventory. Computational resources were employed to model the time series of product family demands, covering the period from January 2020 to December 2024. After conducting analyses and adjusting the models based on mean percentage error, the most suitable models were selected, including ARIMA, SARIMA, Holt-Winters, Additive Seasonal Exponential Smoothing, and Additive and Multiplicative Damped Trend Seasonal Smoothing. The demand forecasts served as the basis for simulating and optimizing scenarios aimed at reducing inventory costs. The results indicated that an inventory availability factor of 98.78% during the replenishment lead time reduced total inventory costs by approximately 45%, based on the availability factor (z) in the deterministic optimization model, leading to a service level of 99.79%.