The design and use of Performance Measurement Systems (PMS's) for industrial improvement and control have received considerable attention in recent years. Indeed, industrial performances are now defined in terms of numerous and multi-level criteria to be synthesized for overall improvement purposes. Only a few quantitative models for PMS's have been proposed in order to better monitor the continuous improvement cycle. Among them is the one proposed by the authors, based on a fuzzy Choquet integral aggregation operator. It allows expressing an overall performance according to the relative importance of criteria and interactions between elementary expressions. This article is a contribution to the decision-maker's requirements for optimizing the improvement of an overall performance versus the allocated resources. It is proved that with a Choquet Integral aggregation, the resulting profile of elementary performances for optimal improvement can only take particular forms, which greatly aid the diagnosis and control tasks.