So-called ‘real-world’ experimentation is an important part of the innovation process, particularly in the emergence stages. Such ‘real-world’ or in vivo testing of products, technologies and/or processes is thought to generate alternative types of knowledge and data from traditional in vitro laboratory experiments. It is argued that in vivo experiments occur in uncontained settings, are less controllable, and present opportunities for unintended experiences or surprises. In this chapter, we use the example of ongoing experimentation with connected and autonomous vehicle (CAV) in two English urban settings, Oxford and Greenwich, to show how the experimental ‘real world’ is constructed and decided upon by key experimental actors, including some groups (e.g. selected innovators, investors, selected publics) while excluding others (e.g. community groups). These constructions of the ‘real world’, we suggest, may limit space for surprises and second-order learnings as part of the innovation process, which may result in limited learnings from the local projects and restrict contributions to the shared rules and emerging sociotechnical trajectory.