Ex ante evaluation of agricultural research for development projects has become important in recent years for priority setting, ex post impact assessment and learning about generalizability to other populations and contexts. We apply farm household and random utility modeling to baseline survey data and evaluate the impact of a cassava research for development project in Malawi prior to its implementation. The project is being implemented to unlock the potential of cassava in response to the global food crisis. We find that a high proportion of farm households are not self-sufficient in food production and can be assisted by increasing the productivity of land and labor in production, processing and marketing of cassava to reduce deficits and increase marketed surplus. The research for development embeds research in an innovation systems network and speeds up exposure, awareness, adoption and diffusion. This increases the likelihood that incremental benefits will be generated and accrue earlier compared to the counterfactual without the project.