Precision medicine is both exciting and overhyped. There are obvious areas of success, notably in cancer treatment, but much of the promise remains just promise. Big data cannot always help if it is not the right data. A commentary today in JAMA Internal Medicine speaks directly to the problem of this lack of data, which is in large part due to current strategies for running clinical trials. Clinical trials with their well defined data collection and endpoints are the ideal vehicles to supply the data needed for precision medicine. By their nature, however, they are designed to look at a broad group of patients, and focus on major endpoints. Precision medicine looks for subgroups of individuals, but current subgroup analyses from clinical trials are not as helpful as they could be. An accompanying article suggests that suggestive subgroup analyses rarely hold up under further inspection or attempts at reproduction. A big part of this problem lies in the original design of the trial. Their data could be enhanced by careful prespecification of subgroups, stratifying when possible, and being careful about overfitting and other analytical problems that occur when using subgroups. Open clinical trial data access is another important component if we are going to see precision medicine make a real impact.