This paper addresses the question of defining the “real” structure of a city, which accommodates existing activity patterns and facilitates urban management and distribution of the city resources, since this structure does not always correspond to the city administrative division. The motivation for the paper is to present a method of defining spaces of everyday activities (SEAs). Here we define the structure of the city based on user-generated online data, which provides evidence about the usage of the city space in the course of everyday consumer practices. Based on Google Places data for St. Petersburg, we build user-venue and venue-venue networks, where users and venues are connected through check-ins and venues are connected through common users. We develop a clusterization technique to define the units of the geographically proximate and socially similar venues – SEAs. We compare the map of such units with the administrative division of St. Petersburg on two different scales: at the level of agglomeration and at the level of administrative districts. We describe the formation of borders and connections that cause matches and inconsistencies between administrative boundaries and SEAs, such as natural and artificial barriers, transportation and infrastructural connectors.