We problematize this notion of a theoretical failure by exploring the case of the Tea Research Institute of Tanzania (TRIT). TRIT is a national agricultural research institute for the tea sector, and like other similar institutes, has a hybrid governance model that relies upon public and private funding. Our data collection follows a participatory approach where first-hand experience working within the Tanzanian agricultural research system provides the basis for analysis. This is supplemented by interviews with industry actors, NGOs and policy entrepreneurs. Our analysis examines the governance structure of TRIT, the various programs of work and its collaborations with different public and private actors in the tea sector. Through this analysis we highlight where innovations have taken root and spread throughout the tea sector. Specifically, we discuss the emergence of a commercial extension program, the diverse approaches to mechanize harvesting, the plant and soil science laboratory, irrigation technologies, and sustainability certification. These programs illustrate how ‘tinkering’ is done to create space for organizational, production and product innovation (cf. Law and Mol, 2011; Mol, 2008; Mol et al., 2010). We focus our attention on the importance of socio-technical networks and local adaptations of knowledge and technologies. We highlight the challenges and contestations that shape how practices follow markets and preempt policies, but are also reactions to the same. We further argue that the switch from an ‘interventionist’ role to a ‘regulatory’ role in a number of key public agencies has enabled the tea sector to remain vibrant and introduce new technologies into production practices and into the market for tea. Following this analysis, we return to the puzzle of a theoretical failure and we argue that the national level discourses on STI policy remain within discourse while ‘tinkering’ is far more important for the practical applications of STI. The existing agricultural research institutions provide interesting opportunities to mobilize expertise in innovative collaborations with the private sector, which are necessary in order to meet the funding requirements of a research operation in Sub-Saharan Africa. This hybrid model that emerges provides a good example of the ‘innovation policy dance’ (Smits et al., 2010) as theories, policies and practices interact at times in step with the music and at other times off-beat. Indeed the theoretical failure found in this case is actually to be found as a resource for new collaborations.