This matches what the author says no? He argues that clinical trials (i.e. deciding whether to approve a drug) should always be randomized. At the same time he

This matches what the author says no? He argues that clinical trials (i.e. deciding whether to approve a drug) should always be randomized. At the same time he believes there is lots of potential for real-world-data in other steps. Pre-trial simulation, post approval outcome tracking, automated drug-interaction analysis, etc. The data analysis is incredibly valuable, even if not suited for drug approval decisions.