Changes of failure mechanism with increasing confinement, from tensile to shear dominated failure, is widely observed in the rupture of samples in laboratory and in rock masses in situ. However, common failure criteria typically consider only shear mechanisms. A hybrid criteria based on a sigmoid function is introduced to account for a transition from tensile to shear dominated failure with increasing confinement. When evaluated by fitting to an extensive laboratory database the sigmoid criteria does not provide a better fit compared to the Hoek-Brown failure envelope, but provides insight into rock strength controlling factors that have significant consequences with respect to the interpretation of laboratory test results. It also leads to a differentiated approach for design by considering two types of behaviour process: 1) in the inner shell, i.e. the direct vicinity of openings, the failure mode is dominated by tensile cracking leading to spalling and related geometric dilation processes and 2) in the outer shell, i.e. remote from excavations, where confinement promotes interlock, we suggest that rock masses could be significantly stronger than predicted by standard approaches.