Demanda por categoría en retail: integración de eventos comerciales en modelos de pronóstico

Abstract

CONDITION FOR PUBLICATION OF PROJECT. fashion industry is defined by the high variability in demand, highly influenced by external factors such as commercial events, trends, and promotions. In a highly competitive environment, demand forecasting is a key factor for the company’s profitability and sustainability. This investigation aims to develop a forecasting model, incorporating exogenous variables, in order to provide reliable predictions, compared to traditional methods based solely on historical sales data. Through this approach, the aim is to align the estimates with actual market behavior to optimize inventory management strategies. Consequently, it contributes to fulfilling planned goals, mitigating both product shortages and inventory surpluses. Seasonal ARIMA and SARIMAX models were applied to an imputed historical sales dataset to guarantee the reliability of records. The results underscore the critical importance of managing forecasts at the category level, as aggregate forecasts fail to capture the distinct behavioral patterns inherent to each category. Within this framework, SARIMAX models demonstrated superior performance compared to traditional approaches by incorporating external variables that enhanced predictive accuracy. This research concludes that the proposed modeling approach facilitates a more precise alignment of business objectives with actual market demand, thereby optimizing inventory levels and reinforcing strategic decision-making in dynamic and competitive market environments. Keywords: Forecast, Retail, Exogenous variables, SARIMAX, Time series.