This research investigates a global planning problem that integrates production, storage, and transportation decisions. The studied problem occurs in a leading mining industry. The logistics and production system is a capacitated multi-site, multi-level, and multi-product network where products are manufactured, stored, and transported, with time windows constraints, via a single railway to final stocks. To solve this problem, we propose an integrated mathematical model and three heuristic procedures, namely, relax and fix, top-down, and bottom-up. We generate several realistic instances and compare the performance of the heuristics to the exact method. The computational experiments show that the relax and fix heuristic allows good-quality solutions regardless of the considered instance. The top-down and bottom-up approaches reveal some counterintuitive managerial insights related to subsystem trade-offs and interactions. Finally, we demonstrate the high impact of train transport on the performance of the global problem.