Recently, stress-based dilatancy criteria have become essential tools to design underground facilities in salt formations such as gas storage caverns. However, these criteria can depend critically on the volumetric strain measurements used to deduce the dilatancy onset. Results from conventional triaxial compression tests can show different volumetric behavior depending on the loading conditions, as well as on the measurement techniques. In order to obtain a quantitative understanding of this problem, an experimental program was carried out and the testing procedure was investigated numerically under homogeneous and heterogeneous stress states. The experimental results showed that the deviatoric stress corresponding to the dilatancy onset was significantly dependent on the measurement techniques. With a heterogeneous stress state, the simulation results revealed that the strain measurements at different scales (referred to as local, hybrid or global) can provide different volumetric results with moderate to significant deviations from the idealized behavior, and hence different onsets of dilatancy. They also proved that, under low confinement, tensile stresses can take place within the compressed specimen, leading to a great deviation of the dilatancy onset from the idealized behavior. From both experimental and numerical investigations, the difference in sensitivity to the measurement techniques between the deviatoric and the volumetric behaviors is explained by the relatively small values of the volumetric strain. The non-ideal laboratory conditions have more impact on this strain than on the deviatoric one. These findings can have implications for the interpretation of the dilatancy behavior of rock salt, and hence on the geomechanical design aspects in salt formations.