For deep learning models, the selection of samples for annotation has a profound impact on the training results, particularly when working with limited annotation budgets. This problem is challenging and underexplored. Our goal is to identify the most important data for annotation that can be used for transferring shape labels for 3D point clouds. Intuitively, the selected 3D shapes should be representative and cover local and global shape structure variations. Towards this goal, we introduce a stochastic optimization strategy that uses both local and global features to select the most relevant 3D shapes for annotation. Our method is effective, extensive experiments have demonstrated that our method can achieve superior performance compared with the state-of-the-art methods on two 3D shape label transfer tasks, namely 3D shape segmentation label transfer and key point transfer.