This paper summarizes lessons learned from formative research conducted in Bangladesh, Ghana, India, Mali and Nepal to inform the development of newborn health interventions, mostly in the context of field trials. Current practices, constraints to the adoption of optimal practices and implications for implementing inventions to improve newborn survival are discussed for: optimal care during pregnancy; skilled care at birth; optimal delivery and newborn care practices; special care of low birth weight babies; and timely and appropriate care seeking for newborn illness. General lessons concerning target audiences and intervention strategy are also drawn. In brief, interventions to reduce neonatal mortality need to start during pregnancy not only to promote birth preparedness and institutional delivery, but also to start the process of change concerning early newborn care practices. Their target audience should not only be pregnant or recently delivered women, but also include the main gatekeepers, particularly traditional birth attendants, grandmothers and other family members. Health providers' opinions also matter as care practices are less likely to change if families receive conflicting messages from different sources. Interventions are more likely to succeed if they are not simply message based, but include problem solving approaches, and a behavior change component to address community norms. Although antenatal care (ANC) is theoretically a good channel for newborn interventions, capitalising on its potential is not straightforward, and will require considerable investment and intervention development in its own right in order to improve ANC counselling, which will need to extend beyond training and tackle the many working day constraints encountered by ANC providers. Removing or subsidising the cost of deliveries may be a necessary action to increase institutional deliveries, but it is unlikely to be sufficient; measures will need to be put in place to ensure the basic quality of institutional deliveries and newborn care, and to change staff attitudes and practices. Post-natal visits should include observation of the baby, referral and counselling of the mother concerning danger signs in addition to promoting optimal care practices. The lessons learned should guide the development of interventions in other contexts, and ensure that key essential elements are not overlooked. They do not, however, mean that formative research will not be needed in other contexts, although the list of questions to address should be considerably reduced; successful intervention strategies require adaptation to make them local, context-specific if they are to be effective, and ongoing process monitoring to ensure the quality of intervention delivery, to check that it is having its intended effect, and to respond to any concerns from its implementers, recipients or the community. Finally, major gaps in evidence are highlighted. These include: establishing levels of recognition of asphyxiated babies and effectiveness of local solutions for resuscitation; clarifying the extent of the overlap between community perceptions of 'at risk' babies and low birthweight babies; developing and evaluating effective interventions to enable ANC services to deliver effective behaviour change counselling for pregnant and newborn health; evaluating effectiveness of delivering community-based newborn interventions at scale through routine services.