When traversing QCD medium, high $p_\perp$ partons lose energy, which is typically measured by suppression, and also predicted by various energy loss models. A crucial test of different energy loss mechanisms is how the energy loss depends on the length of traversed medium (so-called path-length dependence). The upcoming experimental results will allow to, at least in principle, for the first time, clearly observe how the energy loss changes with the size of the medium, in particular, by comparing already available $Pb+Pb$ measurements with now upcoming $Xe+Xe$ data at the LHC. However, in practice, to actually perform such test, it becomes crucial to chose an optimal observable. With respect to this, a ratio of observed suppression for the two systems may seem a natural (and frequently mentioned) choice. We, however, show that extracting the path-length dependence from this observable would not be possible. We here provide an analytical derivation based on simple scaling arguments, as well as detailed numerical calculations based on our advanced energy loss framework, showing that a different observable is suitable for this purpose. We call this observable path-length sensitive suppression ratio ($R_L^{AB}$) and provide our predictions before experimental data become available. This predictions also clearly show that this observable will allow a simple comparison of the related theoretical models with the experimental data, and consequently to distinguish between different (underlying) energy loss mechanisms, which is in turn crucial for understanding properties of created QCD medium.