<p>Retaining walls are structures that hold soil or other materials behind them. Dry masonry retaining walls are built with non-cemented rock slabs or masonry. Unlike concrete walls, they should be designed to ensure not only general stability, encompassing sliding below the wall, foundation collapse or toppling of the whole wall but rather, designs should also account for intrinsic stability, since these walls can fail due to sliding through masonry contacts or to toppling of parts of the wall. Analytical approaches are used to design these walls, but failures still take place in practice. In this study the authors performed physical models of small-scale walls and compare results against analytical and numerical approaches, with the aim of better understanding the possible failure mechanisms and fine tuning the development of these approaches. Physical models are a sound tool for verifying and analysing some instability mechanisms, and particularly those associated with the stability of dry masonry retaining walls controlled by gravitational forces and frictional strength. The authors have designed and carried out physical model experiments of no backfill walls, which are tilted until failure, and walls backfilled with sand, that are backfilled progressively until failure or final stability. The response of these physical models is compared against analytical calculations showing a good agreement, provided some fine tuning of the calculations is performed. This includes accounting for rounding of the corner of the used granite blocks. Moreover, discrete element numerical models of these small-scale retaining walls are performed, showing a good representation of the mechanisms and stability levels at stake in comparison both to physical models and analytical computations. Granite dry masonry retaining wall failures are still not well-understood. The presented approach shows that some factors such as the rounding of the corners of blocks or the dip of the base plane of potential toppling elements of the wall can be relevant for the occurring of instability mechanisms, so they should be accounted for in design methodologies. Moreover, the study shows that a multi-scope design methodology like the one here proposed (analytical, physical and numerical models) is convenient for the designing of these type walls (new geometries), since it can help to identify issues that can be overlooked when applying only one of the possible approaches.</p>