The interaction between empirical modellers and policy makers is analyzed in terms of the structure of the industry, the products involved, the value chain and its working arrangements. Our account of successful interaction focuses on models as the site for knowledge creation and consensus building in both forecasting and policy analysis. The application of network analysis and ideas from knowledge management enable us to interpret the structures and the institutional arrangements described in a range of case studies. We explain why some institutional arrangements may lead to helpful interaction, but our analysis suggests that the way a structure is used is more critical than the nature of the structure in determining the success of interaction between modellers and policy makers.