It is no news that the strategies and methods of the natural sciences often fail to deliver in the realm of social research. Much of the debate focuses on the way their applicability suffers from the fact that the concepts to be measured are fuzzier, the data is worse, and that the dynamics of society mean the researcher is aiming at a moving target. But more profoundly, a major difference is in the subject matter itself. One significant difference between people and particles is that humans make and do things. They act. A logical conclusion is to not limit legitimate research questions to descriptive What's or Why's, but extend them to the concern of 'What is the best way of doing this?' The activity in question can be as hands-on as assembling a car or as abstract as planning an educational intervention. Such questions necessarily raise normative issues, and it comes as no surprise that some of the first who made them the focus of their research, such as the Soviet psychologist Vygotsky, operated within a Marxist framework. In monitoring and evaluation such a 'best-practice' perspective is adopted as a matter of course, but in fact it can be the guiding principle throughout a study. In educational research many studies take a particular intervention as given and seek to establish its usefulness. Others provide mere inputs into the intervention design, by establishing what factors affect the phenomenon under investigation. Neither tell the whole story, but rarely does an intervention's conception itself serve as the research, even though this requires a rigorous analytical discussion of its own. Not only should this neglect be addressed to link theory and practice, I will argue that some of the most interesting questions indeed require this kind of research to be tackled at all. Contemporary incarnations of such approaches range from explicitly engineering-inspired 'Design Research' to more philosophical (but still practical) 'Activity Theory' á la Engeström. This paper will examine points of convergence between these two particular approaches as well as crucial differences. In particular, it will address the question to what extent they serve to generate theory, whether that is an explicit goal or not.