The long-term objective of the paper aims to provide decision aid support to a technical smart buildings manager to potentially reduce the emission of data produced by sensors inside a building and, more generally, to acquire knowledge on the data produced in the facility. As the first step, the paper proposes to characterize the smart-building ecosystem's Internetof-things (IoT) data sets. The description and the construction of learning models over data sets are crucial in engineering studies to advance critical analysis and serve diverse researchers' communities, such as architects or data scientists. We examine two data sets deployed in one location in the Grenoble area in France. We assume that the building is an autonomic computing system. Thus, the underlying model we deal with is the wellknown MAPE-K methodology introduced by IBM. The paper mainly addresses the analysis component and the adjacent connector component of the MAPE-K model. The content of this layer, and its organization, constitutes the methodological point we put forward. Consequently, we automatically provide a complete set of practices and methods to pass to the planning component of the MAPE-K model. We also sketch a semiautomatic way of reducing the number of measures done by sensors. In the background of our study, we aim to reduce the operational cost of making measures with a much more sober approach than the current one. We also discuss in profound the main findings of our work. Finally, we provide insights and open questions for future outcomes based on our experience.