Previous research has emphasized the importance of shared cognitions such as team mental models (TMM) for high performance in teams. The most prominent approaches to measure TMMs are based on indices of within-group agreement (e.g., r_WG, r_^*WG, and a_WG). However, a major shortcoming of these indices is the missing separation of team-specific agreement. Recent indices overestimate the existence of team-specific mental models if there is a common agreement beyond team borders. In this paper, we discuss team situations in which this team specific agreement is most important (e.g., competitive negotiation or action teams) and introduce the TMM index (r_RG) that is based on random group resampling. The advantage of this index is its focus on team specific agreement rather than on common agreement. We demonstrate the usefulness of the r_RG by comparing it to other agreement indices theoretically and by means of Monte Carlo simulation. Finally, limitations and fields of application are discussed.