Most sophisticated planning and scheduling approaches for the process industry consider a fixed time horizon and assume that all data is given at the time of application. Planning and scheduling approach for a continuous and dynamic decision process where decisions have to be made before all data are available is proposed. As an inspiration we have a real world problem originating from a complex pharmaceutical enterprise. The approach is based on a hierarchically structured moving horizon framework. On each level optimization models are proposed to provide support for the relevant decisions. The levels differ regarding the time scope, aggregation, update rate and availability of data at the time applied. The framework receives input data piece by piece and has to make decisions with only a partial knowledge of the required input. Solution procedures have been developed and the optimization models have been validated and tested with data from the real world problem. The solution procedures were able to obtain good solutions within acceptable computational times. It is believed that multi-scale dynamic online procedures are more suitable than traditional offline procedures for many specific types of planning and scheduling problems found in the process industry and should be explored further. (c) 2006 American Institute of Chemical Engineers.