Brand management on social media A Machine learning based framework using text and images

Abstract

vi newlineABSTRACT newlineEffective brand management on the social media networks has been a challenge for the newlinebrand managers. This thesis focuses on monitoring some of the brand management aspects newlineusing text data from Twitter and image data from Instagram. newlineSocial media increased the expressiveness of users and decreased the firms control newlineover their brand personalities. Brand managers feed massive amounts of firm-generated and newlineuser-generated content to the business information systems to understand the online newlineperceptions of brands. By analyzing 83,264 tweets of 13 prominent Indian brands in four newlinesectors using a machine learning model, we established that some of the newer brands are newlinesavvier in managing their brand personalities than the relatively older ones. The public sector newlinebrands have poorer alignment between communicated and perceived brand personalities than newlinethe private sector brands on Twitter. We provided evidence of positive association between the newlinecommunicated brand personality and brand position. A counterintuitive finding is that the newlinebrand sentiment changes in value but not in category (positive, neutral or negative), after newlineremoving the bots and suspicious Twitter accounts. newlineMaintaining a brand identity via images on social media is a relatively new territory for newlinethe brand managers. The brand identity-brand image alignment on the social media is an newlineimportant yet mostly-overlooked phenomenon. We proposed a scalable Google Cloud Visionbased newlineapproach for measuring the alignment between brand identity and brand image, and newlineunderstanding the brand positions. We analyzed 3247 images of 13 leading Indian brands on newlineInstagram. Images containing wordy announcements by the firms are in stark contrast with the newlinerelatively more emotive images by the users. It leads to a noticeable disconnect between the newlinebrand identity and brand image. Also, the private sector brands do not always outperform the newlinepublic sector brands in branding efforts. By offering practical guidance on how to measure and newlinevii newlinereduce the misalignment, this study pa