DEMAND FORECASTING: AN APPLICATION OF THE HOLT WINTERS METHOD IN A MEDIUM-SIZED TEXTILE INDUSTRY

Abstract

Demand forecasting is a fundamental process for supporting strategic and operational decisions in industrial companies, especially in seasonal sectors such as textiles. This article aims to evaluate the application of the additive Holt-Winters method in a medium-sized textile company located in Santa Catarina, considering forecasting as a dynamic process that requires continuous evaluation, data processing, and periodic adjustments. The research was developed based on a monthly time series of children's clothing production between 2008 and 2011, initially using classical decomposition to identify trends and seasonality, followed by the application of the Holt-Winters model. Forecast quality was measured using the mean absolute percentage error (5.97%) and Theil's U statistic (0.346), indicating satisfactory model adherence. However, outliers were identified in November and December 2010, associated with delays in raw material supply, which distorted the results. After processing these data, performance indicators improved significantly, with a reduction in the mean error to 5.08% and a Theil's U of 0.312. These findings reinforce the importance of understanding forecasting as an iterative process, which depends on both model selection and the quality and consistency of the information used. As a contribution, the study highlights the applicability of the Holt-Winters method to the textile sector, highlights the importance of anomaly treatment, and suggests the future integration of hybrid and digital approaches to increase reliability and support planning practices aligned with Industry 4.0 principles.

DOI: https://doi.org/10.56238/sevened2025.029-110