The reintroduction of propagules gathered on the reference ecosystem can be an excellent alternative to sowing commercial seed mixtures. Hay transfer is now commonly used for the restoration of various grassland-types, particularly those with high diversity but limited species dispersal capacity. Optimization is needed to ensure the success of this method. The main objective of our study is to determine how environmental conditions influence hay transfer success. Our hypothesis is that the higher the competition from established non-target species, the lower the establishment of target species. This competition from non-target species can be lowered by decreasing nutrient content and/or by removing the soil seed bank on degraded sites before hay transfer. We thus synthesized the results from several hay transfer experiments carried out between 2007 and 2013 in south-eastern France. Among these experiments, there were two reference plant communities: Camargue meso-xeric grasslands and La Crau steppe. Experiments were carried out from small-scale (0.16m²) to large scale (20ha). Hay was always gathered by air-vacuum and was transferred on three different substrates: i) previously cultivated, nutrient improved soils, ii) previously cultivated, after topsoil removal or iii) low nutrient content disturbed soil. Our hypothesis was confirmed, the closest communities to the references were those where hay was transferred on soil with no non-target seed bank and low nutrient content.