Active dispersal is driven by extrinsic and intrinsic factors at the three stages of departure, transfer and settlement. Most empirical studies capture only one stage of this complex process, and knowledge of how much can be generalized from one stage to another remains unknown. Here we use genetic assignment tests to reconstruct dispersal across 5 years and 232 habitat patches of a Glanville fritillary butterfly (
Melitaea cinxia
) metapopulation. We link individual dispersal events to weather, landscape structure, size and quality of habitat patches, and individual genotype to identify the factors that influence the three stages of dispersal and post-settlement survival. We found that nearly all tested factors strongly affected departure probabilities, but that the same factors explained very little variation in realized dispersal distances. Surprisingly, we found no effect of dispersal distance on post-settlement survival. Rather, survival was influenced by weather conditions, quality of the natal habitat patch, and a strong interaction between genotype and occupancy status of the settled habitat patch, with more mobile genotypes having higher survival as colonists rather than as immigrants. Our work highlights the multi-causality of dispersal and that some dispersal costs can only be understood by considering extrinsic and intrinsic factors and their interaction across the entire dispersal process.