The soil reinforcement method, which consists of placing a granular mattress above a set of rigid inclusions, is used to transfer part of the surface load to the piles through arching. The addition of an extensible geosynthetic layer at the base of the mattress increases load transfer. A small-scale centrifuge model, the mobile tray device, is used to simulate the foundation behaviour by modelling soft soil settlement between the inclusions. A parametric study was conducted to examine the influence of the mattress thickness and of the pile spacing. The different test configurations are compared with regard to the stress transmitted to soft soil, to differential settlement and to the efficiency of load transfer. With thicker mattresses and/or closer pile meshing, load transfer increases and surface settlements are reduced. Geosynthetic maximum experimental and analytical deflections are also examined. For the analytical study, conducted using the EBGEO German standard, three different distributions of the vertical stress applied to the geosynthetic layer are compared: uniform, triangular and inverse triangular. All three scenario results fall within the range of the experimental results. The inverse triangular distribution reveals some deflection values that were lower than the uniform one. The triangular distribution gives the highest deflection values.