This presentation addresses two criticisms to implementation studies from the third generation. First, they put too much emphasis on “predicting the type of implementation behavior that is likely to occur in the future” (Goggin et al., 1990:18). In contrast, drawing on seminal studies from the first generation of implementation studies (Cohen et al., 1972; Pressman & Wildavsky, 1984), I argue that non-linearity and unpredictability are two essential features of any policy process. Thus, what matters for implementation studies is to differentiate how policy stakeholders react to these unexpected events, and whether their reactions lead to a greater or to a lower fulfillment of policy objectives.My second criticism to third generation studies is their binary vision of policy outcomes which leads to an overlook of how power relations between various stakeholders shape implementation processes. Indeed, Goggin et al. (1990: chap. 7) evaluate the failure or success of policy programs based on the initial objectives set by the legislator, without analyzing who benefits the most from policy successes and policy failures. In other words, Goggin et al. (1990) consider that any public policy pursues the common good, and seems to ignore that getting access to legislative arenas and implementation resources is easier for some actors like corporate businesses, than for others like civic and cultural organizations (see Stone, 1989).To go beyond third generation studies, I draw on two recent studies that have called for more scholarly attention to be paid to a) the various streams of actors and events (Howlett, 2019), and to b) the power dynamics that are shaping policy implementation and its rules in use (Sager & Gofen, 2022). Thus, I apply the Multiple Streams Framework of Kingdon (2003), with a particular attention to the stream of politics. I focus on the impact of five federal policy changes in the planning policies of two Swiss cities (Zurich and Geneva) over a period of four decades (1980-2019).My overall hypothesis is that cantonal authorities are not equal when it comes to reacting to federal policy changes and implementing them. Those benefitting from a long-term vision of planning policies and of more implementation resources will react by creating critical junctures of the three streams (policy, problem and politics) which will foster the implementation of other policy objectives (case of Zurich). In contrast, implementing agents with a short-term vision and less implementation resources will favor solutions aligning with their interests, but undermining the fulfillment of other policy objectives (case of Geneva).