Building LCA aims at guiding designers towards more sustainable projects. Many sources of uncertainties affect environmental modelling. Therefore, the reliability of decisions based on building LCA is questioned. However, information on uncertainties can strengthen decisions, when properly addressed. In this study, seven statistical metrics are compared based on dependent samplings for sensitivity analyses (SA) and uncertainty analyses (UA). It allows to identify a set of indicators on which conclusions are more reliable. For the first time, this methodology is applied to comparative building LCA, considering three construction alternatives: a concrete, a concrete blocks and a wooden-framed house. A new SA method, based on Morris, helps identifying which of 153 uncertain factors are more likely to influence decisions. More precise data is then collected on these uncertain factors. In this case study, the Heijungs significativity metric and the distribution of relative differences were the most appropriate metrics to assess UA results. They allow to determine, for each indicator, which is the best alternative and how much better it performs. Consequently, the non-conclusive indicators are discarded. Applying this methodology, decision are enhanced by uncertainties and rely on a smaller set of indicators to select an alternative in terms of environmental performance.