We propose a new, control theoretic methodology for defining performance measures of mobile manipulators. As a guiding principle, we assume that the kinematics or the dynamics of a mobile manipulator are represented by the end point map of a control system with outputs, and that a locally controllable system yields nontrivial performance measures. In the paper, we focus on two categories of dynamic performance measures: the compliance measure and the admittance measure. In both these categories, the following local and global performance characteristics are introduced: the agility ellipsoid, the agility and mobility, the condition number and the distortion. The usefulness of new local measures is demonstrated on the example of determining optimal motion patterns of a wheeled mobile robot.