Put your money where your mouth is: Using deep learning to identify consumer tribes from word usage

TL;DR

This paper introduces Tribefinder, a deep learning system that automatically identifies virtual tribes on Twitter based on language use, aiding firms and researchers in understanding online consumer groups for strategic marketing.

Contribution

The paper presents Tribefinder, a novel deep learning tool for automatically detecting and analyzing virtual tribes on social media, which was previously not possible.

Findings

Tribefinder successfully identifies three main tribal categories.

Language and social interaction metrics differ across tribes.

The system provides insights for targeted marketing strategies.

Abstract

Internet and social media offer firms novel ways of managing their marketing strategy and gain competitive advantage. The groups of users expressing themselves on the Internet about a particular topic, product, or brand are frequently called a virtual tribe or E-tribe. However, there are no automatic tools for identifying and studying the characteristics of these virtual tribes. Towards this aim, this paper presents Tribefinder, a system to reveal Twitter users' tribal affiliations, by analyzing their tweets and language use. To show the potential of this instrument, we provide an example considering three specific tribal macro-categories: alternative realities, lifestyle, and recreation. In addition, we discuss the different characteristics of each identified tribe, in terms of use of language and social interaction metrics. Tribefinder illustrates the importance of adopting a new lens for studying virtual tribes, which is crucial for firms to properly design their marketing strategy, and for scholars to extend prior marketing research.