Banks occupy a central place in the modern global economy. Their stability is an indicator of the state of the country's economy as a whole, which has been repeatedly confirmed by the history of financial crises. The trend and intensity of global digitalization are radically changing the architecture of markets, which requires the development of new theories, methodologies, the search for new management technologies for immediate response to all kinds of challenges and for making informed decisions based on a comprehensive system assessment. The purpose of the study is to conduct a comparative probabilistic assessment of the performance of banks based on the generation of a random multidimensional value of financial indicators. The hypothesis of the study is that probabilistic multidimensional comparative evaluation models will eliminate subjectivism, increase the efficiency and reliability of the raw data base for generating management decisions. The authors have developed a new methodology for comparative evaluation of banks using multidimensional probabilistic analysis. The main problems of managerial decision-making in conditions of uncertainty are identified, taking into account the presence of an anthropogenic factor in the system, the stages of formation of a training sample of commercial banks are described, a list of statistically significant financial indicators is selected, a mathematical problem is formulated, a methodology and mathematical tools for analyzing multidimensional indicators are defined. Using a practical example for 2015-2020, a training sample of banks was formed, divided into two clusters, the coefficients of the equation of the separating hyperplane were determined, a multidimensional random variable was generated, the probability of banks being assigned to one of the clusters was calculated. The results of the calculations showed that only some banks managed to keep their place in the "positive" cluster and the units showed a positive increase in probability. The scientific and practical significance of the research lies in the increment of knowledge on the development of a methodology for multidimensional probabilistic assessment of the position of banks in the training sample. The basis of this methodology can be extended to related spheres of economic life of society, to form the basis of automation models for assessing the financial condition of the subject, finding solutions to optimization problems and developing management solutions.