In this paper, results obtained from a series of large-scale repeated load model tests are presented. Repeated load tests were conducted on geocell reinforced and unreinforced dense sand layers overlying weak sub-grades to understand the general behavior of reinforced unpaved roads. The height of the weak sub-grade was always maintained and the height of the dense sand layer was varied according to the heights of the geocell mattress used. The relative densities of the weak sub-grade and dense sand were maintained at 30 and 75% respectively. The loading was applied through a circular steel plate which replicates the traffic loading applied through a sophisticated double acting linear dynamic actuator attached to a 3.5 m high reaction frame. Results from each case are presented, and different views on the results are discussed with the experimental tests. It was inferred that the rutting on the pavement surface can be reduced to about 35% by providing nominal size of the geocell mattress in base layers over weak sub-grades. The results are also quantified through a non-dimensional factor, called traffic benefit ratio (TBR), defined as a ratio of number of load cycles applied on reinforced section to the number of load cycles applied on unreinforced section at the same settlement. It was found that the TBR can be increased as high as 15 with geocell reinforcement at a settlement ratio of 5%. © Springer Nature Singapore Pte Ltd 2019.