We reanalyse the global distribution of gullies in order to provide a set of observational constraints that models of gully formation must explain. We validate our results derived from the global data with four detailed case studies. We show that the availability of steep slopes is an essential factor to consider when assessing the spatial distribution and abundance of gullies. When the availability of steep slopes is taken into account, it reveals, with a few exceptions, that gullies are found almost uniformly across the whole 30°–90° latitude band. Our analysis also reveals that massive ice deposits are anti-correlated with gullies, and that the undulations in the equatorwards limits of the gully distribution could be explained by longitudinal variations in maximum surface temperatures (controlled by variations in surface properties, including thermal inertia and albedo). We find a sharp transition in both hemispheres between pole-facing gullies, which extend from 30° to 40°, to a more mixed, but dominantly equator-facing orientation of gullies polewards of 40°. We have no definitive explanation for this transition but, based on previous studies, we suggest it could be linked to the availability of near-surface ice deposits.