Life Cycle Assessment is often considered as an evaluation method the application of which rarely provides enough information to take decisions, mainly because of high uncertainties of data and obtained results. These uncertainties can partly be related to existing variability among processes, which are represented as average processes in LCI databases. However, instead of calculating average data, considering sources of data variability could provide useful information to the LCA user, in order to make LCA more helpful for decision making, e. g. with the objective of improving environmental performances of products. The proposed idea is thus to develop an action-oriented LCA systematic method based on sensitivity analysis. The first step consists in separating parts of the technological system related to different economic actors, i.e. the agricultural system, transformation processes, the use phase, the end of life treatment. In a second step, each part of the system is then analysed in terms of models that are used to calculate inventory flows. Models are analysed to identify relevant parameters. These must be independent and are classified in three categories: (i) technological parameters that are subjected to controllable variations (i.e. action possibilities), (ii) environmental parameter that are subjected to uncontrolled variations, (iii) and modelling parameters that are subjected to choices made by the expert. In a third step, the influence on impact indicators of each independent parameter is assessed using screening and global sensitivity analysis. In a last step, sensitivity results are used to identify action possibilities in order to improve the environmental performances of the product. The method is developed for the example of hemp-based construction materials and focuses on a single economic actor: the agricultural system. Models used to calculate emissions from tractors and fertilization operations are analysed. [no pdf]