Many frequent bike-sharing users check bike availability at the station before their intended pickup time via the service provider’s mobile app or website. It is at this time that the user decides whether or not to visit the bike station based on her perceived likelihood that there will be bikes available by the time she arrives at the station. If the station status shows low bike availability, the user may decide to use an alternative transportation mode without even visiting the bike station (i.e., balk). Having estimates on riders’ balking threshold (in terms of bike availability) and time of balking (time of checking the station status) is critical for service providers as it can help improve rebalancing operations and minimize lost demand. What makes this estimation problem challenging is that data collected on bike pickups do not provide any direct information on users’ balking behavior. This paper proposes a novel parsimonious data analysis method to estimate balking threshold and timing of balking from the observed pickup time data (readily available in virtually any bike-sharing system). Since individuals’ true balking behavior is unobservable, we use a simulation model as a testbed to assess the proposed method. The results confirm that the proposed data analysis method can properly estimate both balking threshold and timing of balking decision.