Strategic Insight from Unstructured Data: Essays on Innovation, Brand Communities, and Entrepreneurship

This dissertation explores how unstructured data, textual and visual content shared by individuals and organizations, can offer valuable insights into strategic priorities, brand communities, and entrepreneurial activity. Leveraging recent advances in artificial intelligence, particularly natural language processing and computer vision, the three essays apply and extend scalable approaches to analyze patent texts, social media posts, and image-based signals. Together, the studies illustrate how multi-modal unstructured data can uncover overlooked patterns in innovation, brand positioning, and entrepreneurship. The first chapter examines whether and when female leadership promotes innovation that targets women’s needs. Using a text-based measure of female-focused innovation constructed from pharmaceutical patent titles and abstracts, the study finds that female representation in top management teams is not sufficient: only under specific personal conditions, such as parenting daughters, do leaders redirect strategic attention toward more inclusive innovation. The second chapter proposes a new framework for analyzing brand communities on Instagram. Combining computer vision and community detection, it identifies lifestyle-based consumer segments based on how logos and hashtags co-occur in millions of posts. The analysis of Adidas and Nike across global cities reveals both shared and localized brand meanings, offering a methodological contribution to brand positioning and community analysis. The third chapter connects narcissistic self-presentation on Instagram to entrepreneurial activity across U.S. cities. Using a deep learning model to identify selfies as a proxy for non-pathological narcissism, the study finds that selfie prevalence is positively associated with startup quantity but not startup quality. The findings contribute to regional entrepreneurship and psychological economics by introducing a scalable, image-based behavioral signal of entrepreneurial orientation. Across these chapters, the dissertation advances the marketing field by showing how multi-modal, unstructured data, encompassing patent texts, social media images, and hashtags, can be utilized to investigate innovation focus, brand communities, and the drivers of entrepreneurship.