Adhesive bonding modelling is often realised using cohesive zone models (CZM). For pure mode I loading, these laws represent the cohesive stress versus the interface displacement evolution designated as traction-separation laws (TSL). They enable the description of the interface irreversible phenomena such as damage and/or plasticity, while permitting a refined evaluation of the cohesive stress along the overlap. However, these laws are usually chosen a priori. For brittle and ductile adhesives the TSL shapes usually chosen are, respectively, bilinear softening and elasto-plastic. But the development of direct CZM measurements has highlighted that the law shapes can be more complex. The wrong initial choice of the TSL shape can then have an impact on the simulation results reliability. In this article, several methods used to evaluate CZM parameters are compared in terms of TSL shape robustness. Synthetic noisy data generated from a trapezoidal CZM are used for the inverse identification of a bilinear softening TSL. By applying this procedure on different types of synthetic measurements (respectively, Force-displacement, J-integral, backface strain and DIC) the ability of these techniques to capture parameters for a chosen CZM shape that is not the right one enables a rigorous evaluation the robustness to the law shape.