In hierarchical planning systems, up to three different kinds of abstraction hierarchies can be used: consistent aggregation and disaggregation are based on aggregation hierarchies. By representation hierarchies, general rules can be represented, and different levels of granularity can be integrated consistently. Finally, problem—solving hierarchies are exploited to solve complex search problems efficiently. While structure and properties of its components impose fixed, distinct levels for aggregation and representation hierarchies (e.g. a “type”, “family”, and “item” level), problem—solving hierarchies are derived from user—defined criticalities of its components and, therefore, are continuous. In order to combine efficient support of interactive, hierarchical search for solutions of complex planning problems with overall consistency control, a database—oriented planning system should comprise not only fixed—level aggregation and representation hierarchies, but also continuous problem—solving hierarchies. In this paper, traditional approaches to computer- based production planning and hierarchical planning are examined with regard to their abstraction capabilities. Based on that evaluation, a hierarchical search concept for supporting complex planning decisions is proposed. Design guidelines, application examples, derivation methods, and consistency requirements for all three types of abstraction hierarchies are presented. Conceptual modeling and implementation of an appropriate database-oriented, hierarchical planning system are discussed. Being an application of our concept to a make-to-order and make-to-stock balancing problem in mechanical engineering, the PLASIM system is presented.