Abstract We developed an experimental method to investigate organization design and grouping decisions more specifically. We demonstrate the method in a study with 285 participants. The participants were asked to group a set of nine roles into units using card-sorting. The role descriptions indicated that there were interdependencies between some of the roles. Participants’ grouping decisions were quantified and compared against an algorithmic solution that minimized coordination costs. It was found that a relatively small difference in task complexity between groups greatly affected participants’ performance. We discuss how the method can be extended to study a range of variables related to decision-making about organization design.