Machine tool's productivity is a function of the dynamic response between the spindle nose and table, which varies as a function of drive positions within the machine work volume. The position-dependent structural dynamics results in varying stability of the machine. This paper presents a computationally efficient methodology to evaluate and improve dynamic performance of a machine tool at the design stage. An efficient position-dependent multibody dynamic model of a machine tool is developed based on reduced model substructural synthesis. The experimentally validated reduced machine model simulates position-dependent behavior with significantly less computational effort than commonly used full order Finite Element models. The proposed modeling strategy is used to identify weak components of an experimental machine, which limit the productivity due to chatter. The identified weak machine component is modified and the complete dynamics are rapidly analyzed by virtually re-assembling the machine using reduced order models. Optimal design modifications are shown to increase productivity by ∼25%. The proposed method can be used for efficient simulation of structural dynamics, stability assessment as well as interactions of the CNC and cutting process with the machine tool structure in a virtual environment.