Data-driven research, or the science of letting data tell us what we are looking for, is in many areas, the only viable approach to research. In some domains like adaptive clinical trials and emerging research areas such as social computing, useful results are highly dependent on the ability to observe and interactively explore large volumes of real datasets. Database management is the science of efficiently storing and retrieving data. Data mining is the science of discovering hidden correlations in data. Interactive data-driven research is a natural meeting point that presents a new research opportunity. The ability to conduct effective data-driven research requires to combine efficient indexing and querying from databases and pattern mining and classification from data mining to help analysts understand what lies behind large data volumes. In this paper, we explore key challenges and new opportunities in building robust systems for interactive data-driven research.