The success of social robots, even (or, especially) in use cases as simple as "manning" a booth to promote a product in a shopping mall, depends on their ability of interacting with humans in a timely, effective and enjoyable way. In this paper we present an attention system and a gesture system for use by an autonomous social robot in applications related to product promotion. Our attention system employs a modular approach and attention functions, to allow for rich run-time behaviours to arise from simple rules, while the proposed gesture system allows for tuning the gestures to convey different emotions and robot personalities. The two systems were tested in an experiment involving 790 participants, aiming to explore the attractive power of different robot behaviours.