In the context of the EU-funded JOLISAA (JOint Learning in Innovation Systems in African Agriculture) project, four local innovation processes involving smallholders in Benin were selected for in-depth assessment: innovation in hwedo agrofishing, integrated soil fertility management (ISFM), rice parboiling and soy value chains. Stakeholders directly involved in the innovation process were interviewed. As alternative to the currently challenged transfer-of-technology model of agricultural research and development (ARD) and in order to develop a framework for process monitoring, this paper presents underlying explanatory/analytical models derived out of lessons learnt from the four case studies, reflecting the main features of and the main drivers behind these innovation processes. The explanatory models are: i) feedback socio-ecological model; ii) pull-and-push innovation-driver model; iii) co-learning and joint selection model; and iv) historical account using the innovation bundle. Each model emphasises specific linkages within complex processes of innovation by smallholders. The models are not mutually exclusive and should be embedded in a more comprehensive theory-based framework of innovation processes. However, each of them can be used to design operational monitoring and systematic ongoing assessment of innovation in focusing on specific aspects of the innovation process. The innovation-bundle model helps in following innovation sequences over time and making local innovation visible. The co-learning model describes how innovations are identified or designed, adjusted and kept, or dropped. The pull-and-push model helps project designers think of innovation in terms of opportunities for smallholders and not in terms of problems. The feedback socio-ecological model stresses the relationship between changes in the state of natural ecological systems and human actions and institutions. Monitoring would contribute to making the innovation process more visible and p