Well-designed experiments can help teams understand causal effects, but they’re often expensive to implement. As such, interest in data fusion – whereby observational and experimental data are combined – has increased. In this paper, Rosenman and Owen propose using observational data to influence the design of an experiment, thereby preserving the unbiasedness while improving the precision and efficiency of randomized controlled trials. Specifically, they use data from observational studies to determine allocations of units to strata and treatment assignments.