In courses with hundreds of students, online or hybrid implementation may become more practical than standard classroom teaching. However, it can be difficult for teachers to track student progress in all areas reliably in large courses. In this paper, we present a study where two large online computer science courses were analyzed. Detailed data about student performance in different types of exercises and assignments were collected. In addition, students' perceptions about their learning performance, and the quality and difficulty level of learning materials were collected during all seven weeks of the course. The performance data was analyzed to try to recognize the effectiveness and quality of different course areas. Moreover, we found out if the time usage or perceived difficulty level affected students' performance. The strong correlations between different types of exercises and exam scores indicate that the material is effective and the exam measures the learning properly. However, time usage and perceived difficulty level seem to have little effect on the result.