We continuously face high dropout rates in our faculty's Bachelor degree programs in Computer Science and in Information Systems. This is a quite common problem across engineering and computer science departments. Economically, for the university dropout means a waste of resources. On the other hand, to the young people it means a waste of lifetime, even though they might gain some experience. Furthermore, the labor market demands for many more well-educated engineers and IT specialists than are currently output. When we want to take special care of students that are in danger of dropping out, we must ensure that we identify those that really need this support. This group is sometimes called the “murky middle”. To identify them, we not only have to differentiate them from those students that will eventually earn their degree even without counseling, but also have to distinguish them from those students that do not seriously study at all - which is quite a significant number. Data analysis techniques on students' demographic and progress data have shown potential for an early prediction of dropout. We use logistic regression to identify factors for study success or failure, and we apply decision tree analysis to support the process of deciding which students we should invite to special counseling programs. Our models reach a similar high accuracy as previously reported studies with respect to predicting whether a student will drop out after the first semester. We also demonstrate that we are able to predict reasonably well whom we should - or should not - invite for counseling after the first semester.