While recognizing the need to match an evaluation design with the specific question at hand, randomized controlled trials are often considered the gold standard in producing unbiased estimates of effect size to determine ‘what works.’ (See, for example, federal government websites such as the Department of Education’s What Works Clearinghouse and the Department of Justice’s Model Programs Guide.) Still, there can be political, ethical, and practical barriers to effectively implementing a randomized design. Robert Boruch, among others, has written on the questions to be asked before a randomized trial is initiated. The Vermont Mobile Polling Project was a prime example of a program able to be evaluated using a randomized design. Ultimately, it still presented a number of challenges to effective randomization. Studying cases of randomized trials in action, such as the Vermont project, may improve evaluation implementation in future studies by first recognizing, and then beginning a dialogue, on how to address threats to internal validity in actual cases.#Evaluation2009 #Cluster,Multi-SiteandMulti-LevelEval #2009Conference