BACKGROUND: Embedded implementation research (IR) promotes evidence-informed policy and practices by involving decision-makers and program implementers in research activities that focus on understanding and solving existing implementation challenges. Although embedded IR has been conducted in multiple settings by different organizations, there are limited experiences of embedded IR in humanitarian settings. This study highlights some of the key challenges of conducting embedded IR in a humanitarian setting based on our experience with the Rohingya refugee population in Cox's Bazar, Bangladesh. METHODS AND FINDINGS: We collected qualitative data in between January and July 2019. First, we visited Rohingya refugee camps and interviewed representatives from different humanitarian organizations. Second, we conducted interviews with researchers from BRAC University who were engaged with data collection and analysis in a broader embedded IR study on maternal, newborn, child, and adolescent health (MNCAH) program implementation challenges. Data were analyzed using a thematic analysis approach. Two researchers developed and agreed on codes and relevant themes based on the objectives of this study. The findings of this study highlight several challenges encountered while conducting embedded IR in the Rohingya emergency setting in Cox's Bazar, which may have implications for other humanitarian settings. The overall context of the camps was complex, with more than 100 organizations devoted to providing health services for approximately 1 million refugees. Despite the presence of the Bangladesh government, United Nations agencies and other international organizations played key roles in making programmatic and policy decisions for the Rohingya. Because health service delivery modalities and policies and related implementation challenges for MNCAH programs for the refugees changed rapidly, the embedded IR approach used was flexible and able to adapt to changes identified, with research questions and methods modified accordingly. Access to the camps, reaching Rohingya respondents, overcoming language barriers in order to get quality information, and the limited availability of local research collaborators were additional challenges. Working with researchers or research institutes that are familiar with the context and have experience in conducting implementation and health systems research can help with collection of quality data, identifying key stakeholders and bringing them on board to ensure the execution of the project, and ensuring utilization of the research findings. Study limitations include possible constraints in generalizing our conclusions to other humanitarian settings. Implementation research conducted in additional humanitarian settings can contribute to the evidence on this topic. CONCLUSIONS: Findings indicate that embedded IR can be done effectively in humanitarian settings if the challenges are anticipated, and appropriate strategies and in-country partners put in place to address or mitigate them, before commencing the funding or starting of the project. Understanding the context and analyzing the role of relevant stakeholders prior to conducting the research, considering a simple descriptive method appropriate to answering real-time IR questions, and working with local researchers or research institutes with specific skill sets and prior experience conducting research in humanization contexts may reduce costs and time spent, and ensure collection of quality data relevant for policy and practice.