Many forecasts are guided by so-called “patient allocation” exercises. Here, prescribers are asked to predict how many of their patients they would treat with each current product and (usually) one novel product—whose profile they are shown. This seems a sensible approach. However, you must distinguish between research questions (what you want to know) and survey questions (what you ask respondents to tell you). It is usually the case that turning your research questions directly into survey questions produces sub-optimal results. In this case, you want to know how many patients will be treated with your new product and you directly ask for that information as a patient allocation exercise. In this brief article, we explain why this does not work as well as you might hope.