Designing a Data Mart for Bank Risks and Creating a System of Data Quality Checks

This paper contains an architectural approach to developing a Data Mart for banking risks with a methodology for creating a data quality assessment system for this data mart. The paper also analyzes the retail risk department of a particular large bank, to which the obtained approaches to developing a data mart with several data sources and evaluating data quality are applied. The main problem in developing a data mart is the presence of several data sources, which requires architectural solutions specific to data warehouses, as well as the need for quick access to data from the department, which leads to the need of choosing a hybrid approach with elements of atomic and multidimensional architectures. Another big problem for the bank is the assessment of the data quality in this data mart and the data quality itself. An approach is needed that will cover the data mart with inspections not only of a technical nature, but also verifying the business sense, but at the moment there is no approach that allows this to be done in the field of banking risks. The approach described in this paper to the grouping and hierarchy of checks, both in the technical and logical parts of data quality, allows us to solve the problem of a long assessment of data quality and automate most of the checks based on standardized groups. To design a data mart architecture and draw up an approach to assessing data quality, an analysis of the bank's retail risks was carried out in accordance with the financial services data model, an analysis of internal and external reporting, as well as data requirements from scoring models, was carried out. Also, the architecture of the data mart was developed, and checks were drawn up to the data mart to assess its quality and the requirements for a system for monitoring the quality of the storefront data were identified. An analysis of the results from the implementation of the data mart and the data quality assessment system was also carried out, and satisfactory results were obtained showing the feasibility of the developed approaches for implementation in banking risks. To solve the above problems, domestic and foreign sources were used, for example, the model of the six primary dimensions for assessing the quality of data by Nikola Asham, a leading coach and expert in data management, materials for designing data warehouses and data marts by Ralph Kimball and Bill Inomon, as well as their modern followers , and the experience of Russian banks in the design and testing of warehouses by Tinkoff and Sberbank, and materials on the criteria for assessing the quality of data by Nelson R.R. and Amrapali Z.