Running Randomized Evaluations