This paper concerns some aspects of uncertainty in a multi-criteria decision-making (MCDM) in the framework of recommendation systems. We try to emphasize the interest of handling uncertainty with possibility distributions in a MCDM process when evaluations are imprecise. We propose to support the recommendation process by managing the propagation of decision-making indicators (location, imprecision) through the multi-criteria aggregation. Marginal contributions of partial criteria to the aggregated results w.r.t. these indicators are the foundations of our recommendation principles.