Using robust optimization for distribution and inventory planning for a large pulp producer

• Modeling of a large distribution and inventory problem for a large pulp producer.
• Development of a flexible robust solution approach.
• Robust solution is much better than deterministic approach with safety stock.
• Large industrial case from Södra Cell is used for computational experiments.
• Simulations and analysis show large improvement over manual solutions.

Södra Cell is a world leading producer of pulp and has a large distribution network for its pulp products. This network includes production mills in Sweden and Norway, terminals in European harbours and inland locations, and many customers. The company uses several transport modes including long chartered vessels, spot vessels, trains, barges and trucks. The company uses a supplier managed inventory with a large proportion of its customers. This makes the logistic planning including transportation and inventory critical, as Södra Cell has direct responsibility of their customers’ inventories. However, there is still some uncertainty regarding customer demand and Södra Cell has traditionally used a safety stock inventory approach to handle this. In this paper, we introduce a robust optimization approach to handle the uncertainty and to establish a distribution plan, together with related inventory management. With this approach there is no need for explicit safety stock levels. This is instead taken into account directly through the robust solution. In the proposed model, we can use practical characterization and historical information on the uncertainty. An important result with this is that we can avoid solutions that are too conservative and costly as in standard robust models. A large case study from Södra Cell is used to validate the proposed approach against the traditional approach with safety stock. The analysis is based on simulations and it shows that the robust approach is more cost efficient and can be used as a basis in a decision support system.