As multiphysics geomechanical models are developed, their increasing complexity and number of parameters make it particularly difficult to calibrate against experimental data. In this contribution, I present a heuristic workflow to invert for parameters of a coupled Thermo-Hydro-Mechanical (THM) model in a way that helps the theoretical modellers refine their definition of the underlying elasto-visco-plastic model itself. I apply this workflow to the calibration of deviatoric and volumetric data for two sets of triaxial experiments on mudstone and sandstone. I show that beyond the calibration of well-defined parameters of the THM model, one or two scaling factors of the experimentally obtained yield surface and two more factors to capture the confinement and pore pressure dependency of the flow law lead to the satisfactory matching of two series of six experiments at varying confinements for two different rock types. Using this physical model, I also show that, for the problem of volumetric pore collapse, tracking the volumetric component of the mechanical power in the numerical simulations might allow reducing the number of experiments required to calibrate the model.