Modern cities, industrial plants, cars, trucks, and vessels, among others, make extensive use of cyber-physical systems and sensors. These systems are very critical and contribute to assist decision making. Large data streams are thus produced and analyzed to extract information that allows building knowledge through a set of principles called wisdom. However, because of multiple imperfections, as well as intrinsic, contextual, and extrinsic conditions that alter data, the quality of the generated streams must be evaluated, to determine how relevant they are for decision support. This paper presents a methodology to monitor cyber-physical systems by quality estimation, which defines suitable evaluation characteristics for pertinent analysis. Quality assessment is defined for data imperfections, information dimensions, knowledge factors, and wisdom aspects. The case study of a cyber-physical network of a liquid container training platform is presented in detail, to show how the approach can be applied. Obtained measures are multidimensional, heterogeneous, and variable.