Robot elastostatic calibration facilitates high-accuracy positioning with high payload. Stiffness identification is an important step in this. Poses and forces/moments chosen for stiffness identification determine compensation quality because they influence the propagation of errors impacting stiffness identification to compensation errors. For predefined applications, poses and forces/moments for stiffness identification that maximize positioning accuracy must be selected. Also, two error sources influence stiffness identification, namely, deflection measurement uncertainty and errors in forces/moments applied. Both these error sources’ impact on compensation quality must be minimized. This paper introduces a framework to choose poses and forces/moments for stiffness identification which minimizes above mentioned error sources’ impact on compensation quality. It also maximizes accuracy after compensation at any pose(s), along any axe(s) and with any load(s) that the specified application demands. This framework is applicable for non-over-constrained robots in which considering compliance only along active joints is sufficient, like for most serial-robots and hexapods. Its efficacy was validated using simulated and experimental elastostatic calibrations of a bipod and a high-precision positioning hexapod, respectively. Using this framework to optimize robot geometric calibration is also discussed.