The increasing complexity of technology development and adoption is rapidly changing the effectiveness of scientific and technological policies. Complex technologies are developed and disseminated by networks of agents. The impact of these networks depends on the assets they command, their learning routines, the socio-economic environment in which they operate and their history.

This paper argues that impact assessment research has not made more of a difference because the measurement of the economic impact has poor diagnostic power. In particular it fails to provide research managers with critical institutional lessons concerning ways of improving research and innovation as a process. Paper's contention is that the linear input-output assumptions of economic assessment need to be complemented by an analytical framework that recognizes systems of reflexive, learning interactions and their location in, and relationship with, their institutional context.

This paper reviews a recent donor-funded project concerning the introduction of post-harvest technology to poor hill farmers in India. Rather than conform to conventional development aid projects of either a “research” or an “interventionist” nature, it combines both approaches in a research-action program, which has more in common with a business development approach than a formal social science one. An important conclusion is that the work (and apparent success) of the project is consistent with an understanding of development that emphasizes the importance of innovation systems.