Scholars have invoked subgroups in a number of theories related to teams, yet certain tensions in the literature remain unresolved. In this article, we address 2 of these tensions, both relating to how subgroups are configured in work teams: (a) whether teams perform better with a greater number of subgroups and (b) whether teams perform better when they have imbalanced subgroups (majorities and minorities are present) or balanced subgroups (subgroups are of equal size). We predict that the impact of the number and balance of subgroups depends on the type of subgroup—whether subgroups are formed according to social identity (i.e., identity-based subgroups) or information processing (i.e., knowledge-based subgroups). We first propose that teams are more adversely affected by 2 identity-based subgroups than by any other number, yet the uniquely negative impact of a 2-subgroup configuration is not apparent for knowledge-based subgroups. Instead, a larger number of knowledge-based subgroups is beneficial for performance, such that 2 subgroups is worse for performance when compared with 3 or more subgroups but better for performance when compared with no subgroups or 1 subgroup. Second, we argue that teams perform better when identity-based subgroups are imbalanced yet knowledge-based subgroups are balanced. We also suggest that there are interactive effects between the number and balance of subgroups—however, the nature of this interaction depends on the type of subgroup. To test these predictions, we developed and validated an algorithm that measures the configurational properties of subgroups in organizational work teams. Results of a field study of 326 work teams from a multinational organization support our predictions.