Until recently, food Life Cycle Inventory (LCI) data were relatively rare. As more food LCI data sets are being released, we enter a new phase where, at the same time data coverage remains to be improved, but also data consistency and database harmonization become real challenges so that users can access LCI data covering the large diversity of agricultural systems and products. In this article, we discuss how LCI databases should be built, considering the different kind of uses they serve. We argue that the databases that are more ecodesign/upstream oriented such as AGRIBALYSE, will go towards increasing modelling accuracy and complexity, whereas other databases that are more focused on downstream users and labelling schemes may prefer more simple approaches and models, and easier repeatability. We propose a solution that would enable combining these data with different scales and accuracy, so that harmonization does not lead to overall lower quality for users. This scheme could be used for instance in projects such as the Product Environmental Footprint (PEF).