With student and staff wellbeing a growing concern, several authors have asked whether existing data might help institutions provide better support. By analogy with the established field of Learning Analytics, this might involve identifying causes of stress, improving access to information for those who need it, suggesting options, providing rapid feedback, even early warning of problems. But just investigating the possibility of such uses can create significant risks for individuals: feelings of creepiness or surveillance making wellbeing worse, inappropriate data visibility destroying trust, assessments or interventions becoming self-fulfilling prophecies. To help institutions decide whether and how to explore this area, and to reassure individuals that this is being done safely, we propose a Wellbeing Analytics Code of Practice. This starts from an existing Learning Analytics Code, confirms that its concerns and mitigations remain relevant, and adds additional safeguards and tools for the wellbeing context. These are derived from a detailed analysis of European and UK data protection law, extracting all rules and safeguards mentioned in relation to health data. We also develop context-specific tools for managing risk and evaluating data sources. Early feedback suggests that these documents will indeed increase confidence that this important area can be safely explored.