When students select a university, one of the important parameters is the average class size. This average is usually estimated as an arithmetic average of all the class sizes. However, it has been recently shown that to more adequately describe students' perception of a class size, it makes more sense to average not over classes, but over all students -- which leads to a different characteristics of the average class size. In this paper, we analyze which characteristic is most adequate from the viewpoint of efficient learning. Somewhat surprisingly, it turns out that the arithmetic average is the most adequate way to describe the average student's gain due to a smaller class size. However, if we want to describe the effect of deviations from the average class size on the teaching effectiveness, then, instead of the standard deviation of the class size, a more complex characteristic is most appropriate.