An information system not providing a confidence degree in the proposed information is in danger of not being properly used. The traditional procedure in evaluating the performances of a system is to consider it as a black box and to analyze its results to different stimulus. Unfortunately, in a changing environment and/or in the case of scalable systems (e.g. updates, new processing techniques available) the previous analyze is no longer reliable. To overcome this problem, we propose a new methodology based on the decomposition of the information system in its elementary modules. In order to be able to directly evaluate the system performances, we make use of two quality views. The first one, local quality, describes the information in the input and the output of each module. The second one, global quality, describes the information proposed to the decision maker. The idea behind this double level quality evaluation is that the local quality (much easier to be evaluated than the global one) can be used to directly estimate the global quality. For passing from local quality to global quality, we introduce the notion of quality transfer function. This function is analogous with the transfer function from signal processing domain and it is particular to each processing module. It has the role of capturing the module influence over the information quality. Characterizing a processing module by the means of a quality transfer function allows a direct connection between the input quality and the output quality. Employing in practice our methodology allows having an instantaneous quality estimation of the information proposed by the information system.