Validation is often considered the cornerstone of expert systems evaluation. Validating expert systems has, however, turned out to be a difficult task because an expert system is often both a piece of software and a model. The general procedures for validation of traditional software cannot, therefore, usually be followed when validating an expert system. This article shows how the validation of a concrete full-scale expert system for financial statement planning was conducted, utilizing guidelines on validation of expert systems given in literature and incorporating them into the classical spiral model used in developing the system.