Methodological and data-analysis challenges for evaluators working in disaster and crisis contexts are discussed. Crises are, by definition, unexpected, nonroutine occurrences that create conditions unfavorable to traditional methods of data collection. The inherently novel nature of disasters and large-scale crises, coupled with their unpredictability, often makes data collection difficult at best. Given the methodological limitations imposed by the conditions surrounding disasters, researchers and evaluators are often criticized for methodological decisions concerning data collection, randomization, and generalizability. This chapter addresses issues of data collection, randomization and data analysis in disaster research, outlining the difficulties of randomization, problems stemming from the absence of randomization, and potential solutions to these problems.