This article discusses a controversy that arose out of a grassland experiment in the Netherlands. Using the same data, one group of farmers and scientists concluded that a newly developed trajectory towards sustainability in dairy farming was highly effective, whilst a second group of scientists linked to the Research Institute for Animal Husbandry (PR) concluded the opposite. This article seeks to disentangle this controversy and, in so doing, discerns three levels of discussion. The first regards the understanding of agricultural processes of production as constantly changing practices. Here the concepts of co-production and novelties are introduced. The second level regards the methods for research design and analysis. Thirdly, there is the level of institutionalized research routines. These routines come down, amongst other things, to more or less standardized research questions, hypotheses and methods. Basically, level three contains a specific, and necessarily narrow, selection of concepts and methods from the first and second levels. The question, though, is whether such a selection is in line with markedly changing practices in agriculture. The article concludes that institutionalized research routines are unable to represent, understand and support novel and promising practices correctly.