In this work, we present our prediction results for 17 psychological scales based on online data from the Russian social network VKontakte. As part of this project, we predicted a variety of psychological traits, starting with personality traits from the OCEAN inventory and ending with Raven's Matrices. We then introduced 2 models based on the source of digital data: the first consisting of semi-structured social profiles and the second of public page subscriptions. These models cover most public non-textual data on VKontakte. Lastly, we applied this model to real data with a view to construct psychographics for the audiences of different brands.