Conducting a numerical simulation for cohesive granular mixtures that is well comparable with experiments has always been a challenge. In this study, a systematic methodology is proposed for increasing the fidelity of Discrete Element Method (DEM) simulations of cohesive powder mixtures. Segregation of granules during heap formation of a ternary powder mixture is simulated as a proof of concept. The mixture contains three model particles, one of which is an enzyme placebo granule (EP), in order to simulate the segregation of the actual enzyme granules used in detergent formulations. These granules are at a low content level (similar to 2 wt%) and are highly prone to segregation. In this study, the segregation tendency of the EP granules is mitigated by coating them with Polyethylene Glycol 400 (PEG 400). The resulting adhesion is expressed in terms of equivalent interfacial energy for the DEM numerical simulations and is tuned by careful calibration using the concept of the angle of repose. The Cohesion number is used to scale material stiffness or changing the particle size for faster simulation. The particles shapes in DEM are modelled as clumped spheres based on the X-ray tomograms of the real particles. The rest of the DEM input parameters are also selected and tuned based on the particles physical and mechanical properties. Considerable reduction in segregation tendency of the low-level ingredient granules is observed as a result of coating its surfaces by PEG400. Following the proposed calibration strategy, the DEM simulations have predicted the experimental trends closely and a reasonable agreement is achieved. It is observed that using the Cohesion number for scaling the interfacial energy can significantly reduce the number of calibration trials. (C) 2018 Elsevier B.V. All rights reserved.