We run a laboratory experiment to investigate how the size of the group affects coordination in a bank-run game played repeatedly by participants facing different fellow depositors. For comparability purposes, we keep the coordination tightness constant across different sizes. Participants exhibit an adaptive behavior, since the main drivers of their decisions to withdraw are: previous-round outcomes and own initial choice. Moreover, they mainly adopt the best response to previous-round feedback. However, a sizeable share of participants adopts the opposite mode of behavior, that we refer to as experimentation. The analysis of the determinants of experimentation suggest that subjects adopt this behavior when the probability to lead the group toward the efficient outcome is higher. Finally, our analysis shows that the size of the bank has a significant effect on participants’ decisions, since they withdraw more and experiment less in large banks.