The Critical Role of Unstructured Data in Modern Marketing Research
(A2026-133097)

AUTHORS

Reza (Pouria) Khansari, Ivey Business School at Western University; Kersi Antia, Ivey Business School at Western University; Abhishek Borah, Professor of Marketing, INSEAD; Khaled Boughanmi, Columbia University; Ishita Chakraborty, UW Madison

ABSTRACT

The proliferation of unstructured data—whether questions and answers, reviews, music, or speech—has given marketing researchers unprecedented access to subtleties of human expression that were previously impossible to study at scale. Recent advances in natural language processing and deep learning now allow scholars to extract meaningful insights from words, voice tonality, and music acoustic features. This special session comprises four studies that illustrate how modern computational tools uncover latent meaning from unstructured data: from brand mentions in song lyrics to evolving user preferences on streaming platforms, AI-augmented customer Q&A systems, and the dynamics of speech in customer-employee service interactions. Together, these papers demonstrate how unstructured data may be leveraged to advance empirical and theoretical frontiers of marketing research. The first paper by Borah, Rhodes, V. Kübler, Cascio Rizzo, and Banerjee investigates whether brand mentions in song lyrics—known as Brand Name Dropping (BND)—, especially luxury BND, help or hinder a song’s popularity and why. The study answers these questions by utilizing a multimethod approach combining large-scale field data on 2,966 songs from the Billboard Hot 100 and Spotify with controlled experimentation. The second paper presented in this session, coauthored by Yin, Boughanmi, and Mukherjee, introduces a Transformer-based recommendation system that learns user preferences dynamically from limited interactions during consumption. Incorporating acoustic features of songs, their meta-learning framework demonstrates superior predictive performance compared with other state-of-the-art models. In the third presentation, You, Chakraborty, and Arora differentiate user-generated content (UGC) on online platforms that is generated pre- (Question Answering (QA)) or post-purchase and leverage this distinction to propose a novel generative QA model. Drawing on Amazon QAs and reviews, it develops a retrieval-augmented-generation framework with answerability detection and human-centered evaluation to assess the quality and trustworthiness of AI-generated responses. In the fourth and final presentation, Khansari, Damavandi, and Antia delve into the three understudied aspects of speech-based communication in customer complaint handling interactions. Analyzing more than 30,000 conversational turns from 1069 recorded service calls, their study models the dyadic exchange and trajectory of verbal and vocal empathy between customers and employees and links them to immediate (satisfaction) and longer-term (churn) outcomes. By offering insights based on fresh perspectives and innovative methodologies, our proposed session is likely to be of significant interest to academics interested in digital and service marketing, as well as those interested in how AI and unstructured data are reshaping marketing theory and practice.