Unlike Quality by Testing approach, where products were tested only after drug manufacturing, Quality by Design (QbD) is a proactive control quality paradigm, which handles risks from the early development steps. In QbD, regression models built from experimental data are used to predict a risk mapping called Design Space in which the developers can identify values of critical input factors leading to acceptable probabilities to meet the efficacy and safety specifications for the expected product. These empirical models are often limited to quantitative responses. Moreover, in practice the smallness and incompleteness of datasets degrade the quality of predictions. In this study, a Bayesian approach including variable selection, parameter estimation and model quality assessment is proposed and assessed using a real case study devoted to the development of a Cationic Nano-Lipid Structures for siRNA Transfection. Two original model structures are also included to describe both binary and percentage response variables. The results confirm the practical relevance and applicability of the Bayesian implementation of the QbD analysis.