Experimentation, once confined to laboratory and scientific activities, is now seen as one of the essential methods of urban collective action and a privileged mode of urban transformation (Laurent, Pontille, 2018). However, today, many local governments are wondering how to go beyond the individual experiment, how to integrate it into a broader innovative approach that is both idiosyncratic and sustainable (Evans et al., 2016). In this respect, the perspective of sustainability transitions is enlightening because it allows us to consider the production of urban innovations according to a trajectory dynamic in which “niche experiments” participate in the reconfiguration of the dominant “socio-technical regimes” (Geels, Shot, 2007). It also shows that experimenting is not innovating and that the temporality, framework and governance of experimentation do not overlap exactly with those of urban innovation. In this proposal, we would like to show that experimentation is in fact the first stage of what we call “urban innovation regimes”. Following Schon (1983), we will consider urban planners as "reflective practitioners" capable of implementing three distinct forms of experimentation: the demonstrative experimentation (test), the exploratory experimentation (try) and the transformative experimentation (transform). These generate collective learnings (Evans et al., 2021), development lineages and methods of governance which, being amplified (Lam et al., 2020), generate three types of urban innovation regimes. If the first two (the urban regimes of demonstration and exploration) describe situations of governance through innovation, the third (the urban regime of transformation) raises the question of the governance of innovations from civil society. For each of the three identified regimes, we would like to present ideal-typical cities: Grenoble (demonstration), Nantes (exploration) and Madrid (transformation).