This paper develops a multistage stochastic programming to optimally solve the distribution problem of refined products. The stochastic model relies on a time series analysis, as well as on a scenario tree analysis, in order to effectively deal and represent uncertainty in oil price and demand. The ARIMA methodology is explored to study the time series of the random parameters aiming to provide their future outcomes, which are then used in the scenario-based approach. As the designed methodology leads to a large scale optimization problem, a scenario reduction approach is employed to compress the problem size and improve its computational performance. A real-world example motivates the case study, based on the downstream oil supply chain of mainland Portugal, which is used to validate the applicability of the stochastic model. The results explicitly indicate the performance of the designed approach in tackling large and complex problems, where uncertainty is present. (C) 2017 Elsevier Ltd. All rights reserved.