This paper conducts in the frame of the model development Sustainability Impact Assessment Tool (SIAT) a critical analysis at science-policy interface. The analysis emphasises the methods and the process of end-user involvement that aimed at surveying model requirements. We reveal potential problems of institutional embedding of the decision support system SIAT and give recommendations to avoid shortcomings while developing the model design. The meta-model SIAT is the central tool of the European research project SENSOR, which has been developed in the period from 2005 to 2009. First, we illustrate the methodology of the SIAT that is tailored to simulate land use policies. SIAT allows conducting ex-ante sustainability impact assessment towards the target year 2025 at the level of 570 European regions. Then, the critical analysis at the policy-science interface discusses the procedure of the SIAT development process and reveals the mean of prototyping as basis for the requirement analysis. We summarise the major problems we faced at intuitional level that influenced the quality of the requirement analysis. Finally we conclude on the institutional reasons for asymmetric information that the discontinuous and inhomogeneous environment of our target institution (i) hinders efficient stakeholder involvement and (ii) causes shortcoming to mirror precise end-user requirements in the architectural model design. Quantifying end-user utilities on realistic needs is not precisely applicable due to (a) high opportunity costs to survey and harmonise individual requirements, (b) uncertain forecasts on costs estimates, (c) asymmetric information related to high transactions costs for communication and strategic behaviour of policy makers, researcher and IT developer, (d) requested but unfeasible technical implementation possibilities, (e) predefined and thus limited 'room of manoeuvre' and constraints laid down in research proposals and resulting contracts. We conclude on experience-based potential shortcoming in the project design and during the model development process and give final recommendations on a potential strategy to avoid them.