Digital profiling is a datamining technology that finds patterns of behavior in large amount of data, with the aim of forecasting future events by correlating data traces about past behavior. Profiling is being used by corporations and regulatory bodies across multiple domains such as security, tax, finance, and health. What is a profile and what does it say about the subject it purposes to capture? The paper analyses the commonalities and differences of various profiling practices, exposing their specific rationale: prediction, targeting, personalization. Investigating the type of governmentality algorithmic profiling is producing, the paper investigates the background rationality of profiles: subjects reduced to their traceable behavior, to correlation and repetition of itemized data. A subject at the same time hyper-contextualized and radically decontextualized, a hypervigilant citizen expected to constantly conform to an ever-changing norm.