Plant community assembly on recently degraded or newly restored ecosystem can be described through the hierarchical filter model. Understanding the functioning of key filters (dispersal, abiotic and biotic filters) is a major prerequisite for successful restoration. The model includes a global species pool, divided in the internal species pool already present on site (i.e. seed bank) and the external species pool, which may disperse to the site (i.e. seed rain). Seed bank studies are thus carried out to identify the resilience potential on degraded sites and topsoil transfer potential efficacy. However, because of scale issues (differences between sampling methods and size of seed bank studies vs. large scale restoration) and of differences in growing conditions (greenhouse studies vs. on-site), seed bank studies can be misleading for restoration. We show examples of reference ecosystem seed bank found to be scarce and poor in target species (4 sp. on average; mainly Brachypodium distachyon and Poa bulbosa) while topsoil transfer allowed the transfer of ca. 60 species / 4 m². Seed rain studies are complementary to seed bank studies to identify resilience issues. Sites with rare seedling establishment and virtually inexistent seed banks can be considered as requiring propagule reintroduction to promote their restoration. If seed rain studies show that some seeds are transported to these degraded sites, the restoration process should then not only focus on propagule reintroduction but mainly on germination and establishment as crucial filters.