This paper describes how function-means modeling techniques and spatial coupling analysis can be used in early design stages to increase robustness and knowledge about geometrical sensitivity of an assembly design. By incorporating spatial constraint decomposition in function-means modeling, overall spatial constraints may be broken down into nominal dimensions and tolerances for geometrical features. The analysis is carried out at the very early design stages, before CAD models are created and when only simple sketches exist. By documenting all information about functional requirements, design parameters and constraints in a hierarchically decomposed function-means structure, spatial couplings and overall sensitivities may be detected and alternative, less sensitive, concept solutions may be developed. Analyses of functional and constraint couplings increases the knowledge of and understanding for the geometrical sensitivity of a complex assembly design. This knowledge and understanding then serves as a base for tolerance allocation, where tolerances are to be allocated in the most functional and cost effective way.