In this paper we describe a public display system that detects the users' interest and adapts the on-screen content accordingly. An interest estimation algorithm based on the analysis of the users' non-verbal behaviour, including the users' position, their orientation and the social context, is proposed. A preliminary field study suggests that an adaptive public display may be more appealing than a control condition, where the same content is offered without any adaptation. We argue that behavioural-based measures are valuable data to inform and adapt a public display in a social-aware way, improving the users' engagement.