What conceptual platform guides a realistic evaluation of programs aimed to change behavior? Change is difficult and happens slowly: the subjects involved vary between hesitation, enthusiasm, reflection, comings and goings. This essay presents a model that is contrary to the mainstream tendency of treating programs as independent units of linear development. To evaluate program results it is necessary to discover the mechanisms that influence people’s behavior: these are instead often «invisible» because tacit, too familiar, and therefore taken for granted (including waiting, announcing, filters in selection, and exit expectations). Such mechanisms can be understood by studying the phases, the variations in the implementation, the neglect and «blind alleys». Decision makers should also take into consideration these mechanisms and build programs that reflect paths to change instead of jumping-off points.