Résumé : Experimental findings in the literature reveal that DP steels with identical strength and composition, but different microstructures, exhibit inconsistent trends under either necking-controlled or damage-controlled ductility tests. A special case of this phenomenon is referred to in the forming community as the cut-edge failure or edge cracking issue of DP steels. It is observed that globally ductile microstructures are prone to premature damage at cut edges while a comparatively less ductile DP steel performs better under the locally applied deformation at such edges. In this paper, a systematical statistical study is conducted to analyse this paradox from a micromechanical point of view. The obtained results qualitatively confirm the experimental observations and allowed us to explain/rationalize them, as follows. To reach a higher necking-controlled ductility, i.e. global ductility, the mechanical contrast between ferrite and martensite must be increased. Since martensite reveals a very high initial hardening and fast saturation, activation of martensite plasticity will lead to early necking in the microstructure and should therefore be avoided; a high mechanical phase contrast has this effect. However, at the same time a higher mechanical phase contrast is detrimental for the local ductility. More heterogeneity in the microstructure causes the generation of higher local plastic strains and triaxial stresses which leads to early damages. Therefore, in DP steel, increasing the global ductility may come at the expense of decreasing local ductility, and vice versa. The martensite hardening behavior is the main factor governing this phenomenon, and therefore also constitutes the key to processing improved steels that combine the best of both.