Privacy-by-design attribution model: a relative weight and spillover effect approach

• Ano de defesa: 2023
• Autor(a) principal: Cardoso, Eduardo Ogawa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Relative Weight; Atribuição; Attribution; Efeito Spillover; Multi-Channel; Multicanal; Privacidade; Privacy; Spillover Effect
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/12/12139/tde-11042024-191247/

This thesis presents the Privacy-by-design Attribution Model (PBDAM), a novel approach to multi-channel attribution modeling that addresses the challenges of capturing hidden interactions and ensuring equitable credit assignment while maintaining consumer privacy. PBDAM\'s performance is rigorously assessed through a comprehensive examination of both synthetic and actual empirical data, underscoring its superiority over conventional multi-channel attribution methods. Through analyzing a synthetic dataset with aggregated daily data and evaluating actual campaign data from major companies in Brazil, PBDAM demonstrates its ability to effectively capture hidden interaction information, generate fair attribution values, and provide valuable insights without compromising consumer privacy. The findings of this research establish PBDAM as a superior model in terms of performance, equitable credit assignment, and privacy- consciousness.