Abstract. Despite various methodologies proposed in the last decades, there are still no standard for estimating landslide hazard. Consequently, practical applications for territorial management have to assimilate in a single cartography information obtained at local level with different methods, with negative consequences on the quality of derived products. Here we proposed a new methodology – based on well-established hazard matrices – to assess landslide hazard, which starts from a landslide inventory, and introduces a new method for estimating the landslides frequency. We apply this new method to three landslide inventories compiled with increasing detail. They are: (i) a basic-historical inventory, (ii) a generational-historical inventory (a detailed version of a simple historical inventory), (iii) and a composite multi-temporal inventory (which includes the generational-historical inventory plus the multi-temporal inventory). Results are then compared each other, and to independent measures from Persistent Scatterer Interferometry. Our results highlight the importance to base landslide hazard analysis on a generational-historical inventory that adequately characterizes the complexity of landslide clusters, whereas indicate that multi-temporal mapping is not decisive for the purpose. Overall, our procedure puts landslide mapping back at the center of the hazard assessment chain, raising questions on the reliability and availability of landslide inventory maps.