Agricultural commodity research has very site-specific productivity impacts. Crucial determinants of the magnitude and distribution of research benefits include agroecological conditions for technology generation and adoption, as well as commodity market-structure. This paper presents a process for ex-ante research evaluation which accounts for these factors with a dynamic, spatial multi-market model. Simulation results based on sorghum research in Kenya demonstrate that potential research benefits can vary dramatically across program research target zones. In Kenya, however, population-induced demand growth not technological development will have the greatest influence on future sorghum markets. © 1997 Elsevier Science B.V.