Machine Learning and Unstructured Data Analytics for Digital Marketing

The internet has been a fertile ground for content creation, with people generating vast amounts of data daily. The proliferation of unstructured data can hold significant potential for practitioners and managers, as it may reflect the public's interests and behaviors. Analyzing this data, however, presents challenges for traditional marketing and business researchers. The emergence and evolvement of machine learning (ML), data science (DS), and sophisticated computational tools has given rise to novel approaches that can address these challenges. This dissertation investigates three prominent ML and DS applications within marketing research, demonstrating their capacity to effectively process and interpret complex unstructured data—including images and videos—for business insights and consumer welfare. The first study explores the integration of object detection algorithms with eye-tracking technology to dissect the consumer shopping experience online and on mobile platforms. This combination of techniques offers fresh perspectives on consumers' visual attention, yielding practical insights for digital advertising strategies. The second study employs image segmentation alongside Bayesian analysis to explore skin-tone representation in brand visuals on social media. The proposed approach refines the strategic management of marketing communications. In the final study, I apply rigorous data science methods to scrutinize user engagement with advertising on a short-video social platform. Informed by processing fluency theory, this analysis reveals nuanced patterns of consumer behavior under various contexts, providing essential insights to enhance algorithms for ad ranking and recommendation systems. Collectively, these studies demonstrate the significant capacity of ML and engineering techniques to transform marketing research by offering deeper, data-driven insights into unstructured data within the digital landscape.