We report the strategies and outcomes of requirements analysis in the SAMS project, which aims to detect early signs of dementia from records of user behaviour derived from monitoring home computer activity. Initial requirements elicitation specified a complex monitoring system in the genre of awareness requirements. As the complexity of the requirements for interpreting user activity became apparent, we revised our strategy towards an iterative refinement driven by analysis of recorded data. We propose that analysis of complex awareness requirements may necessitate a tool-driven approach where data mining software is configured to discover new requirements. Requirements for data mining and data analysis in SAMS are described. Application of the conceptual architecture for directing requirements analysis strategy in complex system with human in the loop for data analysis and adaptation is investigated. The implications for the extended view for RE are discussed with reference to systems with potentially unknown requirements. A research road map is outlined for the convergence of data mining and RE.