The case of a brominated RNA crystal structure determination in which standard three-wavelength MAD phasing was unsuccessful because of fast X-ray-induced debromination was reinvestigated [Ennifar et al. ( 2002), Acta Cryst. D58, 1262 - 1268]. It was found that if the data are kept unmerged and if a dose-stamp is associated with each reflection measurement, dose-dependent occupancies can be refined for the Br atoms. Such a parametrization has been implemented in the macromolecular phasing program SHARP. Re ning such dose-dependent occupancies on an unmerged data set gave a dramatic improvement, even for SAD phases from only the first wavelength (peak), and resulted in a good electron-density map after solvent flattening. The adverse effect of radiation damage has been turned into a beneficial one. The crucial difference is made by the use of unmerged data: phasing power is generated through the intensity differences of symmetry-related reflections recorded at different doses, i.e. corresponding to different states of the X-ray-induced debromination. This approach should prove useful in all situations of experimental phasing where site-specific radiation damage occurs unavoidably and undesirably and not only in cases in which radiation damage is purposely being created in order to demonstrate its potential usefulness.