The purpose of this work is to evaluate the efficiency of several calibration methods applied to a six-axis industrial (serial) robot. Specifically, the absolute position accuracy of a Fanuc LR Mate 200 iC industrial robot is improved using two calibration models. The first model is purely kinematic, and takes into account all geometric parameters. The second model considers, in addition, five compliance parameters related to the stiffness in joints 2, 3, 4, 5, and 6. For both models, the so-called calibration (or identification) robot configurations are selected based on an observability analysis. For each model, the efficiency of five different observability indices are compared. The parameter identification is based on the forward kinematic approach, where only the residual of the calibration positions is minimized. The data collection process is completely automated, by using Matlab, a FARO laser tracker, communication modules, and a special end-effector with three optical targets (SMRs). The comparisons of the observability indices and of the two calibration models are made through an exhaustive experimentation (7000 measurements). Results show that the robot's mean position error is reduced from 0.622 mm (before calibration) to 0.250 mm, when using the kinemetic model, and to 0.142 mm, when using the complete calibration model.