The paper proposes an extended definition of "big data" as an interdisciplinary and relative category that combines the properties of data arrays and streams, the requirements for their processing, and the criteria for the practical value of the results. It is shown that the evolution of approaches from the "volume-speed-variety" triad to the inclusion of aspects of quality, variability, and reproducibility is driven by the complexity of sources (sensors, transactions, and multimedia) and the growing institutional requirements. For the financial sector of Russia, three key blocks of application are described: asset management, risk management and work with the client base, where the integration of heterogeneous sources, streaming and batch processing, standardization and verification of indicators provide an increase in the accuracy of forecasts, a reduction in response time and an improvement in the client experience. The issues of using the technology of "big data" in the financial analysis of the enterprise activity are considered. A three-component definition framework (data properties, processing requirements, and utility criteria) and a three-stage methodology (unified data model, processing pipelines, and modeling and verification) are proposed. The article also addresses quality and ethical considerations, including data origin, bias control, model validation, and documentation of transformations. The authors conclude that institutionalized and technologically supported data frameworks create verifiable management value and a foundation for long-term sustainability in financial decision-making.