In problem-based learning large cohorts of students are divided into smaller groups that pursue learning objectives with separate instructors called tutors. This presents challenges for tutors tasked with providing similar educational experiences and assessment of multiple groups of students. Here we evaluated between-group variation in test scores that are attributable solely to the random sampling without replacement process used to form smaller groups. We then compared this with the actual between-group variation in test scores in a university-level zoology class over 4 years. We found the variation attributable exclusively to group formation accounted for a 14.4-16.2 point differential between groups whereas differences in empirical test scores attributable to group formation and other factors such as tutor capacity and group dynamics ranged from 12-18 points and rarely exceeded the variation inherent solely to group formation. This implies ad-hoc strategies for reducing variation between groups at the assessment phase will have limited success.