In this paper we present a methodology to analyse input use in the agricultural sector. The novelty of the theoretical model explained is that it has been developed considering a multi-criteria environment. Thus, the optimal input use condition is determined by the assessment of "multi-attribute utility" and "multi-attribute marginal utility". We show how the approach adopted in this paper is a generalization of the single-attribute expected utility theory. The theoretical model developed is further implemented in an empirical application that studies water for irrigation use as a particular case. Results show how multi-attribute utility functions elicited for a sample of 52 irrigators explain differences on irrigation water use in relative homogenous agricultural systems, albeit exhibiting similar water partial utility functions. We conclude that these differences come from the dissimilar weights that farmers attached to each attribute in the aggregate utility function. The irrigated area considered as case study is located in North-western Spain.