The relative impacts on food purchases of many alternative front-of-pack nutritional labelling systems were tested, with various methods—from opinion pool to nationwide experiments. Clearly, some systems induce better purchasing responses, having better nutritional impacts on food baskets. Nonetheless, we still ignore what the ingredients of an efficient label are. Here, we propose guidance for label designers. To do so, we first propose a typology that breaks down established labelling systems into four elementary components: Directiveness, Scope and Gradation, Set of Reference and Sign. On this basis, we then build seven alternative generic labelling systems that we test in a framed-field experiment enabling us to measure the effect of each component on food purchases in isolation. Our results show that an effective front-of-pack labelling system should be Food-Directive (instead of Diet-Directive) and be displayed on both healthy and unhealthy food. The reference set, which is across categories or within categories, produces the same average nutrition score but generates contrasting behavioural responses.