An important byproduct of the emergence of social networking platforms is an access to abundance of social data in all forms: blogs, clicks, facebook feeds, transactions and tweets. It is of great interest to process this large volume of highly unstructured data to facilitate business decisions, public policy making or better social living. The key challenge lies in the fact that even though data is large in volume, the information content is very limited. Therefore, extracting meaningful answers has become a challenging computational and statistical task. In this talk, I will discuss how to resolve it successfully for important questions arising in the context of crowd-sourcing, ranking and viral advertising. The key to our success lies in the identification of the appropriate statistical framework for the problems at hand.