The importance of the initial packing of a rock cumulus on its flow process and emplacement mechanism has been highlighted by several small-scale experiments (Manzella and Labiouse, 2009) where thousands of terracotta bricks were either randomly settled as a loose material or orderly piled one on top of the other before releasing them on an inclined slope. When bricks were piled, longer runout were observed compared to tests run with loose bricks. The reason of this difference has been highlighted using a 2D Finite Element-Discrete Element code by explicitly accounting for the shape of the blocks and the interactions between them. When bricks are piled, the mass has originally an ordinate structure that tends to be preserved during the downhill motion and only after the slope break it shatters, whereas in the case of bricks randomly settled into the box, the mass behaves as a loose material from the start and more energy is lost from the beginning through both friction and collisions at the base and within the granular mass. When the slope break is smoother, the relatively-coherent structure of the block cumulus is even less disaggregated, as a consequence less energy is dissipated at the toe, the mass can travel further and it preserves the inherited geometries. Simulations confirm the experimental results, giving a better insight on the understanding of the effect of the initial block packing on the longitudinal spreading and on the mechanisms underneath the process of rock avalanche propagation.