Abstract
Modern brand management requires an understanding and use of new types of data such as text, and video data as well as techniques such as deep learning. Given this background, my thesis aims to build a modern data-driven approach to help brand managers successfully manage brands. My dissertation has three essays. In the first, I show how managers can develop metrics from Twitter to monitor brand preference at the micro-temporal and micro-geographic levels. I find that my metric leads online sales and can even predict elections. In the second, I show how managers can ascertain the effectiveness of Facebook video ads by using two measure of video ads that I develop: visual and audio saliency. Empirical test of these measures shows that visual saliency, in particular, within the first few seconds of a video ad can predict the degree of social engagement that the video ad will garner. In the final essay, I propose a three-stage framework of association transfer between and an endorser and a brand: pre-, peri-, and post-endorsement. I find that a two-way transfer of association takes place between a brand its endorsers: while the brand borrows associations, the endorsers’ associations also change. These findings show why brand managers need to consider the endorsement theme and endorsers’ associations when choosing brand endorsers. Taken together, my package of studies contributes to both the practice of and research on brand management.