With the increasing density of powerful personal mobile devices - such as smart phones and tablets - the demand for mobile data connectivity grows rapidly. People rely on accessing their cloud-hosted data and globally available Internet services anytime and anywhere. However, providing ubiquitous connectivity is challenging and sometimes infeasible: Consumed bandwidth is predicted to continue growing exponentially and peaks due to unpredictable flash crowds are especially hard to cope with. Even worse, connectivity is inevitably disrupted when infrastructure breaks in natural or man-made disasters. In addition, connecting remote and developing regions is economically unprofitable, thus aggravating the (increasing) global digital divide.Opportunistic networks are envisioned to mitigate many of these problems. Mobile devices can use their infrastructure-less communication capabilities (e.g., WiFi Ad Hoc, WiFi Direct, Bluetooth), to disseminate data in a peer-to-peer manner. Whenever two devices are within radio transmission range (in contact), this is an opportunity to exchange data and a potential step towards providing a service. Hence, connectivity can be maintained, yet at the cost of increased delay. In making efficient use of contacts, opportunistic networking protocols (e.g., for routing, for content distribution) can benefit from the fact that contacts are not completely random, but structured by human behaviour in a social context. For example, it was shown that identifying clusters of people who see each other often can help find good message relays. Thus, to design efficient solutions, we need a thorough understanding of how human behaviour reflects in contacts (who meets whom, how often and for how long - by chance or intentionally) and methods to translate this understanding into efficient protocols.A promising approach in this direction is to represent the (social) structure in contacts as a contact graph, where nodes represent devices, and edges express how strongly two nodes are connected (e.g., how frequently they are in contact). To solve opportunistic networking problems, we can then use the rich toolset of complex network analysis (CNA) and graph theory to identify the position of a node within the topology of the contact graph and use this to assess a node's utility, e.g., for carrying a message. Typical examples of such tools are centrality metrics, assessing the importance of a node for a certain network function, or communities - clusters of nodes with similar positions within the topology. In this thesis, we put these preliminary ideas on solid ground with three original contributions.