In contemporary industries the demand for very accurate robots is continuously growing. Yet, robot vendors are limited in the achievable accuracy of their robots, as they have no means to provide a direct end-effector feedback. Therefore, most approaches aim to identify an accurate model of the robotic system, thus providing compensation factors to correct the deflections. Models, however, are unable to represent the real physical system in a sufficient manner for path correction. The non-linearities in robotic systems are difficult to model and the dynamics cannot be neglected. A better approach is, therefore, to use direct end-effector position and orientation feedback from an external sensor as, e.g. a Leica laser tracker. The measured data can directly be compared to the nominal data from the path interpolator. Hence, the data are independent of the kinematic robot model. The residual errors can be used to calculate correction values in Cartesian space, which are mapped to each individual robot joint, thus providing a fast path correction algorithm. The evaluation of this procedure has shown that it enables the robot to drive paths at a velocity of 100 mm/s with an RMS of only 0.11 mm. In this paper an algorithm, which is implementing this procedure is presented and evaluated on a iso path.