This article details the challenges we faced in collecting sensitive information in an ethnically and religiously divided community that has experienced recent violence. The discussion is based on a summer 2016 survey experiment we conducted in Jos, Nigeria, to gather information regarding residents' perceptions of local communal violence. We discuss the challenges of such research and our approach to randomized sampling, constructing treatments that minimize the stress to respondents, debriefing to lower the possibility of spreading rumors of conflict, and utilizing computer tablets to increase access to the study for respondents with varying languages and levels of literacy. In particular, we discuss a geographic sampling method used for randomization, which we hope will prove useful to others facing similar randomization challenges.