Designs combining different types of data are increasingly used in educational evaluation, to provide both evidence of impact and an explanation of the processes by which impacts are created. Logic models are visual representations of how an intervention leads via a set of steps from resources and inputs to outputs and then sets of outcomes. Their use has become widespread to underpin evaluations; and they have become of more interest in education as they have been promoted by policy makers and funders including the Education Endowment Foundation (EEF) in England. This paper addresses the question: how can logic models be used to frame and implement educational evaluations using combinations of methods? To do so, the paper draws on theory-based evaluation literature to identify a set of issues to be considered: the role of implementation logic; causal mechanisms; the context of interventions; and the importance of considering and addressing issues around complexity. Using detailed examples from two study designs for EEF evaluations, the paper presents an evidence-informed logic model approach to deal with these issues. The paper concludes by reflecting on the practical and theoretical implications of this approach, laying out a set of key issues to address in future evaluations for which a design framed by an evidence-informed logic model may be appropriate.