Data is not the answer to every problem, and most historical problems require some work beyond the information found in databases. This chapter examines the question of when to stop modeling with data. It works from historian David Hackett Fischer’s account of a “complete” explanation, where Fischer argues that historical explanations are only satisfactory when they have both proved that a discontinuity took place and given some account of why it happened. The chapter investigates the problem of what data-driven analysis can and cannot do through a case study about the treatment of environmentalism in the debates of the US Congress. Text mining can show the explosion of abuse channeled at environmentalists in the 1990s; it can even help us to trace that abuse back to a handful of speakers. But textual data from the Congressional debates does not explain what motivated the speakers or allow us to conjecture whether their attacks were coordinated.