We consider a large-scale scheduling problem in which three multipurpose production facilities in different countries supply a large portfolio of fast moving consumer products to the European market. The operation of each multipurpose plant is characterised by complex internal decisions concerning the allocation of production resources over the week-long planning horizon considered. In addition, there is a large degree of inherent flexibility in the overall system, particularly regarding which warehouses are served by which plant. As a consequence, in order for the most efficient schedule to be found, all three plants must be considered in detail and simultaneously as one large production system. We achieve this by using a very general scheduling representation allowing important features such as equipment change-overs and limited intermediate storage to be modelled. As the size of this fully detailed formulation is beyond the scope of current computing resources, we generate an approximate formulation from it in a formal way by aggregating constraints. Its solution gives a tight upper bound on the production capacity of the system and allows the problem to be decomposed into smaller sub-problems, each involving a single plant. Solving these sub-problems involves finding a detailed schedule separately for each plant using production targets set by the aggregate solution.