The flagship Midwives Service Scheme (MSS) was introduced in 2009 as the first large-scale intervention to address rural retention of midwives in Nigeria. This was a multi-component intervention including financial incentives to midwives, provision of accommodation and systems level support, aiming to improve human resource capacity to provide quality services. This study explores how effectively the scheme’s design and implementation drew on the health system’s context, resources, needs and population preferences, and how it contributed to strengthening health systems at all levels. To meet the objectives, 87 in-depth interviews and eight focus group discussions with policymakers, implementers, midwives and community members were conducted in two Nigerian states and at the federal level. Drawing on a systems-thinking approach, the study developed a new framework examining the fit of the intervention with the local health system’s context considering: i) leadership and commitment ii) policy and financing context iii) human resource management capacity, and iv) stakeholder participation. The framework informed the framing of the study and guided data collection; however, themes were identified and synthesised inductively. The broad principles and features of the scheme were widely supported by program managers and policy-makers across the three health systems levels (local, state, federal). However, its design was based on federal level program managers’ knowledge of maternal health and health worker issues, and limited recognition of the decentralised nature of the health system. Implementation was hampered by inadequate management and logistical capacity to deal with the complex design, poor absorptive capacity of states for the posted midwives, failure to provide continued supervision, and welfare issues that affected the midwives. The MSS was successful in attracting midwives including those employed in the private sector due to the promised pay package and capacity building opportunities offered under it. Several factors affected motivation of midwives and impacted on midwife retention. These include low and unpaid salaries and incentives, housing difficulties and distance of housing from the facility, and travel costs and hardships incurred from commuting to the facility. Unmet career development priorities were an additional source of demotivation. The findings point to poor retention of midwives in both states. Retention was better among retired midwives compared to other categories. Younger midwives were more mobile and exited the scheme mainly to the private sector. The MSS had the potential to bring about system-wide changes; however, weak implementation severely hampered its achievement of the intended outcomes and dampened the expectations of significant improvements in the health systems. The findings underscore the importance of reflecting overall health systems structures and processes and local contextual factors, including local health workers’ preferences in designing effective human resource retention schemes. The scheme is potentially replicable as a bundled package of interventions to improve access to skilled workers in rural communities in LMICs. Since decentralisation critically modifies the decision-making space, an inclusive process where sub-national actors participate in choosing policy options should be a prerequisite.