Andra is conducting scientific experiments in the Meuse/Haute-Marne Underground Laboratory among which REP experiment is a vertical mine-by-test focusing on short and long term hydromechanical response of the argilite to the main shaft sinking. Displacements, strains, and pore pressures will be monitored while the shaft is passing down. Andra and INERIS intend to back-analyse most recorded geomechanical data based on under-excavation numerical technique in order to estimate pre-existing field stresses. The under-excavation interpretative technique consists in determining the pre-existing stress tensor related to a quite large volume of rock based on generalized inversion of geomechanical measurements recorded during the disturbance of the host rock (typically the excavation of an underground opening). In the framework a numerical study aiming to test accurately the sensitivity and numerical stability of this interpretative technique, 3D modelling of a step-by-step vertical mine-by-test, based on REP design, has been undertaken. One major step of the numerical procedure is to calculate each instrument response, coming up with a transient predicted measurement curve versus the progress of the shaft sinking. These important, intermediate results, on which this paper focuses, led the authors to several preliminary operational recommendations, e.g. relocation of sensors, which could not have been determined otherwise. These predictive numerical responses are of importance when a fast but efficient validation of the real data coming out from the field must be performed. More generally, the authors intend to show how a synthetic, 3D numerical conception of a under-excavation interpretative experiment reveals to be unique method to explore different instrumentation layouts and optimize numerous options offered to the engineer in charge of its final design.