Active learning techniques, such as peer instruction and group work, have been gaining a lot of traction in universities. Taking a natural next step in re-evaluating current practices, many institutions recently started experimenting student-centred group exams. In order to assess the feasibility and effectiveness of collaborative assessments, we implemented a study that focuses on lower-level and intermediate-level mathematics courses with large student populations. We collected and analyzed data relating to student material retention, their perceptions of the format, as well as the effects of the group composition (in terms of the strength of its individual members) and question types (multiple choice versus short answer) on the exam success. Although we establish that long-term learning is not strongly influenced by the group exams, students find them useful and enjoyable, with the majority of students preferring group exams to individual ones. Examining group composition and question type results, we conclude that multiple choice questions are more suitable for group exams. We also comment on the format in view of limited people and time resources available for test creation, administration, and marking. Finally, we consider instructor perspectives on the process.