From bats to whales, millions of mammals migrate every year. However, their navigation capacity for accomplishing long-distance movements remains remarkably understudied and lags behind by five decades compared to other animals [1, 2]; partly because, unlike for other taxa such as birds and sea-turtles, no small scale orientation assay has so far been developed. Yet recently, bats became a model to investigate the nature of the cues mammals use for long-range navigation, and surprisingly for nocturnal animals, sunset cues, and in particular polarized light cues, appear crucial to calibrate the magnetic compass system in non-migratory bats [3–5]. This does not appear to hold for a species of migratory bat however [6], and thus the nature of the information used by migratory bats for navigation remains unclear. In this experiment, we asked whether the position of the solar disk per se is relevant for compass orientation in a migratory bat, Pipistrellus pygmaeus. Using a new experimental assay that measures takeoff orientation, we tested the orientation of bats exposed to a shifted sunset azimuth using a mirror at dusk. Bats exposed to a 180° rotated azimuth of the setting sun and released after translocation during the same night shifted their heading direction by ~180° compared to control bats. However, first-year migrants had no clear orientation either as controls or following the same treatment, indicating that unlike birds, these bats do not have an innate migratory direction when released after translocation. This suggests learning is a key component in the long-range navigational system of naïve bats in this species. Our study provides rare empirical evidence for the specific cues and mechanisms migratory mammals use for navigation.