Bend gesture-based input on a flexible device requires a distinct indication of gesture completion. Existing research commonly used threshold-based strategies for discrete input, whereas releasing the bend is used to confirm continuous input. However, the simplest form of these strategies does not allow the rejection of unwanted or unintended input and bi-directional manipulation of the output parameter before confirming continuous input. This work aims to identify user-defined gesture completion strategies with the potential to overcome these drawbacks. Ten blind or low vision (BLV) and fifteen sighted students participated in the study. We identified four and three unique strategies for discrete and continuous inputs, respectively. We believe the reported implications and recommendations with potential application scenarios will help researchers and interaction designers take advantage of these strategies.