<p>In geotechnical engineering, the heterogeneity and variability of soil properties can pose a significant challenge for the design and construction of structures. This often-complex subsurface environment is generally represented by simplified homogeneous subsoil models derived from in-situ soundings and boreholes. However, these models do not consider the natural variability of the subsoil and the related uncertainties and, as such, disregard their influence on the design of geotechnical structures. Geostatistical models can be used to capture the inherent spatial subsoil variability by generating random fields (equiprobable realizations). The Random Finite Element Method (RFEM) is an approach that can be used to investigate the influence of soil variability (random fields) on the design of geotechnical structures. In this study, the technical implementation of the RFEM and its application for a twin tunnel case study in Munich (Germany) were examined. The influence of unexpected geological conditions was conducted within a probabilistic framework with the aim of comparing the results with the deterministic approach (idealized subsoil model). A conventional sensitivity analysis of soil variability was also performed. The results of this study demonstrate that deterministic approaches alone cannot adequately characterize tunneling-induced ground and foundation responses in heterogeneous subsoil conditions. Incorporating subsoil variability through RFEM provides a more realistic assessment, enabling engineers to better quantify risk, identify critical settlement zones, and ultimately improve the reliability of tunnel design.</p>