Production management problems are complex owing to large dimensionality, wide variety of decisions of varying scope, focus and time-horizon, and disturbances. A hierarchical approach to these problems is a way to address this complexity, wherein the global problem is decomposed into a series of top-down sub- problems. We advocate that a single planning architecture cannot be employed for all planning problems. We propose a multi-layer hierarchical decomposition which is dependent on the complexity of the problem, and identify the factors influencing complexity. A systematic stepwise design approach for the construction of the hierarchy and inputs required are presented. The subsequent operation of the hierarchy in an unreliable environment is also explained. Aggregation schemes for model reduction have been developed and blended with a time-scale decomposition of activities to provide the theoretical foundation of the architecture. It is also hoped that this methodology can be applied to other such large-scale complex decision making problems.