Three-dimensional computer-aided design (3D CAD) models with different levels of detail (LOD) are used in various industries for numerous purposes. Therefore, it is necessary to develop techniques to simplify 3D CAD models in order to adjust the LOD of the model according to its purpose. The main purpose of simplification is to minimize the change in the outer shape of the models and to reduce the data size of the models. The key technologies to achieve these purposes are evaluation metrics and simplification operation. Evaluation metrics are employed to select elements to be preserved or removed by calculating the importance of the geometric elements comprising a 3D CAD model. The simplification operation removes the selected elements and fills up the void in the model caused by the removal. Feature volume and type have been the most popular criteria used in evaluation metrics for the simplification of feature-based 3D CAD models. In this study, the concept of feature shape complexity (FSC) is introduced, and a method of adopting FSC as a criterion of evaluation metrics is presented. A prototype system for the simplification of 3D CAD models is then implemented. Finally, the effectiveness of the proposed method is verified by conducting simplification experiments with a complex 3D CAD assembly model.