The aim of this study was to investigate the influence of qualitatively different powder feeder performances on resulting granule size distributions after twin-screw granulation of a highly drug loaded, hydrophobic mixture and a mannitol powder. It was shown that powder feeder related problems usually cannot be identified by trusting in the values given by the feeder. Therefore, a newly developed model for the evaluation of the performance of powder feeders was introduced and it was tried to connect this model to residence time distributions in twin-screw granulation processes. The influence of feeder performances on resulting granule size distributions varied, depending on the applied screw configuration and the used powder. Regarding the hydrophobic and highly drug loaded formulation, which was granulated at an L/S-ratio of 0.5, a pure conveying screw and a medium kneading configuration, consisting of 60° kneading blocks were negatively influenced by poor feeder settings. For optimal settings more narrow distributions could be obtained. For an extensive kneading configuration good and poor settings resulted in mono-modal granule size distributions but were differing in the overall size. Mannitol, a model substance for a liquid sensitive formulation was granulated at an L/S-ratio of 0.075. It was even more important to maintain optimal feeding as mannitol was highly affected by poor feeder performances. Even an extensive kneading configuration could not level the errors in powder feeder performance, resulting in qualitatively different granule size distributions. The results of this study demonstrate the importance of detailed knowledge about applied feeding systems to gain optimal performance in twin-screw granulation.