Public Policies are aimed at the citizens of their respective countries. Instead of the traditional view, of the analysis of public policy in the sense of the State for the citizen, the present project sought to verify the feasibility of creating an information system with a different bias: from the citizen to the State, emphasizing how the citizen is being served by the State in all its aspects. If such a conception is viable, the project sought to answer what the necessary requirements are, how the database architecture should be and whether there would be benefits in this approach, such as, for example, the possibility of detecting fraud in the granting of social benefits. Preponderantly, the control systems of these public policies are developed in a relational model database (SQL). Relational tables have the potential to form powerful queries through joins between them, which, given the significant volume of data, can consume a lot of computational resources, and are usually not available online. This modeling does not allow a complete view of the interaction between the State and the citizen. For several technical reasons, analyzed in the project, the information system needs to be developed in a NoSQL graph database. The present work aims to verify whether a graph database allows a modeling with the citizen as the center of public policies. Through public data from three social income transfer programs in Brazil, a graph was created with 433 million nodes and edges. The study allowed us to conclude that the modeling of public policies x citizens is feasible and, through Knowledge graph, has numerous advantages in relation to the traditional model of relational tables and presents the possibility of fraud detection.