Innopolis University, a newly created Russian IT university, is looking for innovative solutions in the sphere of education. This paper investigates the possibility of creating a prediction model, based on admission metrics. All existing parameters, as well their synthesized combinations, were examined as possible future performance predictors. We demonstrate statistically that Russia's obligatory Unified State Exam does not provide sufficient data for predicting an applicant's academic potential. In addition to examining the existing data, authors conducted an empirical analysis of roots of insufficient correlation between overall high school performance and university performance. As a result, few sociological hypotheses have been formulated and respective changes proposed in the design of the admission process. These new techniques will be tested during the 2017 admission campaign.