Almost all academic institutions regularly use graduate surveys in order to assess the determinants of academic success and labour market chances of their graduates. This paper shows how the information drawn from these surveys can be better used than it has been in most studies so far. It analyses the determinants of academic success using individual, socioeconomic and group information. It therefore combines a broader spectrum of determinants of academic success than usual. For individual characteristics, gender, age, study length, school education background, migration background, and whether the student had a child before graduation are included. In addition, subjective assessment of different skill dimensions acquired during the study period is used. The socio-economic background is not only depicted by the education level of the parents but also by the way the student financed his or her studies. This contribution also makes clear that it is useful to aggregate individual student characteristics at the academic subject level because these group effects have an additional impact on individual achievement and depict selectivity into subjects. In previous studies, only a limited list of group characteristics has been taken into account. In addition, deviations from subject mean grades should be used instead of absolute grades. This controls for idiosyncratic grading in subjects. It is also important to include final school grades in order to control for innate ability and differences in resources students enjoyed before their academic study. Finally, institutional fixed effects should be controlled for in order to eliminate idiosyncratic grading and differences in scope and selectivity in institutions. This paper implies that individual cognitive and written skills and independent work have a positive impact on academic achievement in contrast to teamwork competences such as co-operation, boundary-spanning or oral skills or foreign languages. Also, broad academic skills such as theoretical knowledge problem solving skills or broad basic knowledge do not lead to better grades. The analysis also shows that gender and the academic background of the parents lose their significance when other determinants of academic achievements are included, and that selectivity effects into academic subjects play an important role for the final grades obtained by students. The paper is based on representative data of more than 4,500 graduates from the German state of Bavaria in the academic year 2003/2004.