An extensive experimental study was conducted to investigate the co-effects of surface area and distance between electrodes as well as filler scales on the percolation threshold of piezoresistive cement-stabilised sand. In this route, the electrical resistivity of numerous specimens of different sizes and composed of different content of carbon-based conductive fillers was measured, including carbon nanotubes (CNTs), graphene nanoplatelets (GNPs), and carbon fibres (CFs) with different aspect ratios. In addition, the numerical relations between the electrical percolation threshold and matrix dimensions were expressed for different conductive fillers. Furthermore, the electrical percolation threshold of two large-scale specimens with different shapes (a 10 × 10 × 85 cm3 beam, and a 15 cm size cube) were predicted through numerical relations, and their piezoresistivity performances were investigated under compression cyclic loading (cube) and flexural cyclic loading (beam). The mechanical properties of the specimens were also evaluated. The results showed that the changes in the length, width, and thickness of the matrix surrounded between electrodes had a significant effect on the electrical percolation threshold. However, the effects of length changes on the percolation threshold were greater than the width and thickness changes. Generally, increasing the aspect ratio of the conductive fillers caused a reduction in the electrical percolation threshold of the cementitious geocomposite. The appropriate piezoresistivity response of the large-scale specimens composed of filler content equal to their percolation threshold (obtained by the numerical relation presented in this study) showed the adequacy of the results in terms of threshold dosage prediction and self-sensing geocomposite design. The results of this study addressed a crucial factor for the design of self-sensing composites and pave the way for the development of field-applicable, smart, cementitious geocomposite.