Polymer melts show a strongly bulged rim after loading to the rheometer gap. If such a rim is sheared at Wi approximate to 1 surface particle tracking shows for a monodisperse (M-w = 206 kDa) and a polydisperse polystyrene (PS 158 K) that the surface velocity profile is non-linear from the beginning. For cone angles 0.1 and 0.15 rad recipes are provided how to shape the rim closer to spherical, before the test is started. The recipes are for reptation times around 1 s and have to be time-temperature shifted for longer ones. By increasing the temperature and squeezing the sample to more than the cone-truncation gap, the melt surface tension and the wetting of the tools are used to shape the rim. The efficiency is quantified by measuring the time evolution of the rim curvature prior to shear, the surface velocity profile during shear, and the viscosity and first normal stress difference. The surface velocity profile strongly depends on pre-treatment, but viscosity and first normal stress difference are quite insensitive to it. For PS 158 K the shear rate dependence in the range 0.04 < Wi < 4 has been investigated. Even for the best pre-treatment, a strong departure from the linear surface velocity profile is observed at strains around 10. (C) 2008 The Society of Rheology.