In on-line and especially in off-line programming of industrial robots the attainable accuracy has to be taken into account. Especially in the case of off-line programming along particular trajectories followed, neglecting position errors leads to a need for kinematic calibration procedures which, however, apply to the robot controller level. If end effector error is taken into consideration in off-line programming a compensated commanded trajectory can be programmed. This is different to well-established calibration procedures, because it keeps the original kinematic model of the robot and tries to improve accuracy along the particular trajectory of interest. In this paper, the methodology for measuring, predicting and compensating end effector position errors is presented. A straight line trajectory is used as an example in connection to a particular industrial robotic arm. Measurements are taken using white-light metrology. Based on these measurements an error prediction model is constructed by training an Artificial Neural Network. A second neural network model is trained to yield joint coordinates that minimise position error, which is proved by employing the prediction model on the results of the compensation model.