Recent technological advances have enabled earthquake engineering researchers to develop numerical models of increasing complexity, capable of duly reproducing even the smallest structural detail. In the case of unreinforced masonry (URM) structures, however, because of their discrete and heterogeneous nature, computational performance tends to decrease exponentially as a function of the adopted refinement level, thus confining the applicability of advanced micro-models, according to which each masonry unit is typically modelled separately, to reduced-scale problems. To enable their use at a building scale, and benefit from considering simultaneously out-of-plane failures, local wall-diaphragm interaction and collapses, researchers often need to decrease the level of detail of specific members or sub-structures. In the current literature, however, the influence of the abovementioned simplifications on the quality of micro-modelling predictions has been only marginally investigated so far, while code-based guidelines are missing. To start addressing such knowledge gap, the dynamic response of a shake-table-tested full-scale URM building specimen has been simulated in this work using a very detailed micro-model, and the results obtained were then compared with those of nominally identical models in which, however, the idealisation of some specific structural elements has been purposely simplified. Aimed at further extending the impact of this study, pushover analyses were also performed using the same models. Preliminary outcomes, which may serve as a reference to develop more informed, effective and targeted multi-scale micro-modelling strategies in the future, indicate that: (i) maximum base shear predictions tend to be less impacted by the introduction of modelling simplifications, (ii) despite requiring more labour, the explicit representation of the brickwork pattern generally led to better results in terms of predicted damage propagation, failure mechanisms and displacement capacity, (iii) using equivalent membranes, as opposed to modelling each component of timber diaphragms, provided acceptable results, making it a plausible alternative for practical applications of micro-modelling approaches.