In industrial supply chains, accurate forecasting of demand is crucial to rationalise operations and reduce costs. The study compares four established machine learning models for predicting demand of furniture sales linear regression, decision trees, random forests, and XGBoost using a simulated dataset of furniture sales transactions in 2021–2024.The dataset must contain a variety of elements, including product category, type of material, quantity, geographical location and price information. The ability of each algorithm to predict total sales was evaluated after a thorough pre-processing that included encoding of the data in a categorical manner and normalization of the specific models. According to the results, the ensemble models—primarily Random Forest and XGBoost—demonstrated exceptional accuracy in predicting and successfully identifying complex relationships in the data. The results highlight the importance of advanced ensemble approaches in the supply chain context for demand forecasting, even when linear regression provides a suitable basis and after which the data is processed.