In order to reduce feedback overload, a framework of contention-based feedback was established. However, since a user is randomly chosen, it cannot maximize capacity. Moreover, if a collision occurs, a base station cannot obtain any information to schedule a user; thus, a scheduling outage happens. In this paper, we propose a two-stage opportunistic feedback scheme to maximize capacity with the limited feedback. The proposed scheme guarantees selecting the user with the highest signal-to-noise ratio (SNR) channel state and considers the scheduling outage situation using previous feedback information. In addition, we define the effective capacity to show the relation between the capacity and the feedback overload. Simulation results show that the proposed scheme achieves higher effective capacity with limited feedback than existing contention-based feedback schemes do.