This study addresses the robot that waits for users while they shop. In order to wait, the robot needs to understand which locations are appropriate for waiting. We investigated how people choose locations for waiting, and revealed that they are concerned with 'disturbing pedestrians' and 'disturbing shop activities'. Using these criteria, we developed a classifier of waiting locations. 'Disturbing pedestrians' are estimated from statistics of pedestrian trajectories, which is observed with a human-tracking system based on laser range finders. 'Disturbing shop activities' are estimated based on shop visibility. We evaluated this autonomous waiting behavior in a shopping-assist scenario. The experimental results revealed that users found the autonomous waiting robot chose appropriate waiting locations for waiting more than a robot with random choice or one controlled manually by the user him or herself.