Marketing Analytics Without Personal Identifiers: Federated Learning And DP

The study is devoted to the analysis and conceptual integration of two key classes of privacy-enhancing technologies (PETs), federated learning (FL) and differential privacy (DP), with the aim of developing a holistic framework for solving marketing analytics tasks. The methodological basis of the work relies on a systematic review of current specialized literature and authoritative industry analytical materials, followed by a synthesis of the identified approaches. The obtained results demonstrate that the combination of the decentralized FL architecture with the formal mathematical guarantees of DP creates the conditions for building high-accuracy predictive models applicable to such key tasks as conversion rate estimation, target audience segmentation, and personalization of interactions, while eliminating the need for centralization and direct disclosure of sensitive user data. In conclusion, it is substantiated that the proposed FL-DP framework can be regarded as a technologically robust and ethically sound solution that forms the basis for the transition to a new generation of marketing analytics, despite the persisting significant challenges associated with its practical implementation. The article is intended for data specialists, researchers in the field of machine learning, and professionals involved in the development and implementation of marketing strategies focused on building analytics systems with privacy as a priority.