The influence of social connections on human behaviour has been demonstrated in many occasions. This paper presents the analysis of the dynamic properties of longitudinal (335 days) community data (n=3,375 participants) from an online health promotion program. The community data is unique as it describes how the network has evolved since its inception and because the information exchanged through the network was predominantly about the achievements of participants in the program and therefore influencing behavior through social comparison. The analyses show that the largest component of the community network has characteristics of a small world network. The analyses also show that connections are formed according to a strong attachment preference according to the gender, and a weaker homophily for Body Mass Index. The presented analysis can serve as basis for creating novel interventions that influence physical activity behavior through social connections.