<p>Robust deterioration modelling of geotechnical earthworks requires long-term monitoring at sites that are both statistically representative and operationally capable of sustaining deep instrumentation. Despite its critical influence on model validity, the process of site selection for soil deterioration modelling remains weakly formalised in the literature and is typically treated as a preliminary logistical step. This paper proposes a hierarchical, risk-gated framework for geotechnical site selection that systematically reduces national-scale earthwork inventories to a small number of viable candidates for long-term monitoring. The framework integrates GIS-based macroscopic screening, Multi-Criteria Decision Analysis (MCDA), and targeted geophysical verification within a sequential decision structure. The methodology is illustrated through the selection of a road cut slope from the Irish national road network. Results show that while GIS screening and MCDA efficiently prioritise candidate sites, they are insufficient for final site confirmation in geologically heterogeneous environments. Targeted non-invasive subsurface verification, implemented in this study using Electrical Resistivity Tomography (ERT), proved critical as a binary decision gate, excluding the highest-ranked candidate after identifying shallow bedrock. This geological constraint was incompatible with the primary objective of investigating soil behaviour, as it precluded the installation of deep instrumentation. The proposed framework provides a replicable protocol for mitigating site-selection risk by explicitly bridging the scale gap between national datasets and site-specific subsurface constraints. Although demonstrated in Irish glacial tills, the decision logic is transferable to other settings requiring long-term geotechnical monitoring.</p>