Complex software systems are regularly required to work more, faster and on a broader scale, to adapt dynamically to changing workloads, scenarios and objectives, and to achieve guaranteed levels of Quality of Service (QoS). Satisfying such demanding requirements at design time is a not trivial task, since at runtime the presence of the variability, eterogeneity and non-linear behaviour that characterize complex software systems makes the whole process harder for software designers. Feedback strategies must be introduced to keep track of the QoS status of software systems with the purpose of identifying if such status meets the current requirements and of suggesting, if necessary, the most suitable reconfigurations. In this paper we propose a deep investigation for QoS-based Feedback that makes use of design and runtime knowledge thus to provide a methodology to manage the data over time, while the development advances.