The Geostationary Earth Radiation Budget (GERB) instrument was launched during the 2002 summer together with the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) on board of the Meteosat Second Generation (MSG) satellite. This broadband radiometer will aim to deliver near real-time estimates of the top of the atmosphere (TOA) radiative fluxes at high temporal resolution thanks to the geostationary orbit. To infer these fluxes, a radiance-to-flux conversion needs to be performed on measured radiances. Since we plan to carry out such a conversion by using the angular dependency models (ADMs) developed from the Clouds and the Earth's Radiant Energy System (CERES) experiment, the GERB ground segment will have to rely on some scene identification on SEVIRI data which mimic as close as possible the one from CERES in order to select the proper ADM. In this paper, we briefly present the method we used to retrieve cloud optical depth and cloud fraction on footprints made of several imager pixels. We then compare the retrieval of both features on the same targets using nearly time-simultaneous Meteosat-7 imager and CERES Single Satellite Footprint data. The targets are defined as CERES radiometer footprints. We investigate the possible discrepancies between the two datasets according to surface type and cloud phase and, if they exist, suggest some strategies to homogenize GERB retrievals based on CERES ones.