Abstract A short term scheduling problem from an industrial case is modelled in three different ways. A discretized MILP-model, a heuristic approach and a genetic algorithm are tested. The scheduling task is complicated by a recycle connection. Storage capacity for raw materials and intermediates are restricted. Some productions steps are operated continuously and some in batch mode. A one-week schedule and a two-week schedule are constructed with different models and compared with industrial results. The genetic algorithm outperforms the other approaches, though the computational efforts are larger.