As the number of network services provided in fixed and mobile environments are increasing recently, it becomes harder to expect users to manually pick up appropriate services from all these available. Therefore, it has been proposed that network services are automatically selected in accordance with relational information among people and other social objects, i.e., locations, things and contents. The fusion of online and physical sensing platforms has brought about the possibility of building relational information from sensing data: online logs and sensing data collected at mobile devices could reflect what objects people are related to in their daily lives. The relational information can be represented as a graph in which each node could be a social object and each link could represent a direct relation between two nodes and is called relational metric. We discuss how to build relational metrics from sensing data and propose a system design for it.