Abstract The large availability of data, often from unconventional sources, does not call for a data-driven and theory-free approach to social science. On the contrary, (big) data eventually unveil the complexity of socio-economic relations, which has been too often disregarded in traditional approaches. Consequently, this paradigm shift requires to develop new theories and modelling techniques to handle new types of information. In this chapter, we first tackle emerging challenges about the collection, storage, and processing of data, such as their ownership, privacy, and cybersecurity, but also potential biases and lack of quality. Secondly, we review data modelling techniques which can leverage on the new available information and allow us to analyse relationships at the microlevel both in space and in time. Finally, the complexity of the world revealed by the data and the techniques required to deal with such a complexity establishes a new framework for policy analysis. Policy makers can now rely on positive and quantitative instruments, helpful in understanding both the present scenarios and their future complex developments, although profoundly different from the standard experimental and normative framework. In the conclusion, we recall the preceding efforts required by the policy itself to fully realize the promises of computational social sciences.