Research on new models of care in health service provision is complex, as is the introduction and embedding of such models, and positive research findings are only one factor in whether a new model of care will be implemented. In order to understand why this is the case, research design must not only take account of proposed changes in the clinical encounter, but the organisational context that must sustain and normalise any changed practices. We use two case studies where new models of maternity care were implemented and evaluated via randomised controlled trials (RCTs) to discuss how (or whether) the use of theory might inform implementation and sustainability strategies. The Normalisation Process Model is proposed as a suitable theoretical framework, and a comparison made using the two case studies - one where a theoretical framework was used, the other where it was not. CONTEXT AND APPROACH: In the maternity sector there is considerable debate about which model of care provides the best outcomes for women, while being sustainable in the organisational setting. We explore why a model of maternity care--team midwifery (where women have a small group of midwives providing their care)-- that was implemented and tested in an RCT was not continued after the RCT's conclusion, despite showing the same or better outcomes for women in the intervention group compared with women allocated to usual care. We then discuss the conceptualisation and rationale leading to the use of the 'Normalisation Process Model' as an aid to exploring aspects of implementation of a caseload midwifery model (where women are allocated a primary midwife for their care) that has recently been evaluated by RCT.