Investigating the dynamical evolution of dust grains in proto-planetary discs is a key issue to understand how planets should form. We identify under which conditions dust settling can be constrained by high angular resolution observations at mm wavelengths, and which observational strategies are suited for such studies. Exploring a large range of models, we generate synthetic images of discs with different degrees of dust settling, and simulate high angular resolution (˜0.05-0.3 arcsec) Atacama Large Millimeter/submillimeter Array (ALMA) observations of these synthetic discs. The resulting data sets are then analysed blindly with homogeneous disc models (where dust and gas are totally mixed) and the derived disc parameters are used as tracers of the settling factor. Our dust discs are partially resolved by ALMA and present some specific behaviours on radial and mainly vertical directions, which can be used to quantify the level of settling. We find out that an angular resolution better than or equal to ˜0.1 arcsec (using 2.3 km baselines at 0.8mm) allows us to constrain the dust scale height and flaring index with sufficient precision to unambiguously distinguish between settled and non-settled discs, provided the inclination is close enough to edge-on (i ≥ 75°). Ignoring dust settling and assuming hydrostatic equilibrium when analysing such discs affect the derived dust temperature and the radial dependency of the dust emissivity index. The surface density distribution can also be severely biased at the highest inclinations. However, the derived dust properties remain largely unaffected if the disc scale height is fitted separately. ALMA has the potential to test some of the dust settling mechanisms, but for real discs, deviations from ideal geometry (warps, spiral waves) may provide an ultimate limit on the dust settling detection.