Efficient distribution of products is a critical component of the retail industry, directly impacting customer satisfaction and business profitability. Traditional distribution methods often rely on assumptions and expert manual opinions, leading to defective and error-prone processes. These methods can result in stockouts or overstock issues, negatively affecting customers and businesses. This study addresses these challenges by leveraging advanced machine learning algorithms to tackle the distribution problem. The primary objective of the study research is to increase sales of promotional products by reducing stock shortages and overstocks to prevent unexpected holding costs while ensuring optimization in distribution. By analyzing multidimensional datasets, such as previous promotion data, risk profiles, transactional data, and product similarities, this research aims to develop a system that accurately predicts the future distribution of each product to the allocated stores and provides decision support recommendations. Various machine-learning algorithms were employed during the estimation process. Among them, the XGBoost algorithm, which uses bagging to train multiple decision trees and then combines the results, is chosen as the primary model due to its high accuracy. The implementation of this strategy has the potential to streamline distribution efforts by a factor of 10, ultimately maximizing the effectiveness of delivering promotional products to the most relevant stores. By discovering these patterns, we aim to direct products towards the most efficient distribution channels, thereby enhancing operational efficiency and increasing turnover.