Fonte Viva, a member of the MOL Group, manufactures bottled mineral water at two production plants in Hungary and is one of the country's leading FMCG suppliers. Every production process starts with an accurate demand forecast, and Fonte Viva wanted to make forecasting measurable, predictable, and directly connected to production planning. They contacted LogiNet after reading one of our articles on demand sensing.

LogiNet developed a web application that combines AI-powered demand forecasting with production planning in a single workflow. Built with Python and Django, the platform uses an ensemble machine learning model combining LightGBM, XGBoost, and a neural network to generate weekly demand forecasts for every product and customer. Real-time order, inventory, and product data are synchronised from Microsoft Dynamics 365 Business Central via API.

The project ultimately shifted from make-to-order manufacturing to a stock-driven production model. Forecasted demand determines target inventory levels, which feed directly into a rolling production schedule. A water-filling algorithm balances inventory across products, while a CP-SAT optimisation solver calculates the optimal production plan. Production managers can review and manually adjust the proposed schedule whenever required.

The forecasting model achieves approximately 90% accuracy. Planning has become transparent and continuous across both manufacturing plants, allowing production capacity to be utilised more efficiently without additional investment. The platform also identifies production bottlenecks, helping the business prioritise future process improvements.