No need for costly surveys, Twitter reveals how a brand is perceived

What do Toyota, Aveda, and Clif Bar have in common? A forthcoming article in the INFORMS journal Marketing Science finds that Twitter fans of these brands are all more likely to follow accounts that tweet about the environment. This in turn creates a greener image than other brands in their sectors.

The research, conducted by Aron Culotta and Jennifer Cutler of the Illinois Institute of Technology, examined accounts of millions of fans of over 200 brands across a variety of sectors on Twitter. They find that a brand's image is related to the characteristics of who their fans follow on Twitter.

The authors first searched Twitter lists for accounts that were identified as exemplars of attributes like eco-friendliness, luxury, and nutrition. For example, Greenpeace or Sierra Club are exemplars of eco-friendliness.

Next, they computed a social perception score for each brand on each attribute based on the overlap between each brand's followers and the followers of the exemplar accounts. Interestingly, they find their Twitter perception scores closely match direct survey ratings of brand image on each of the attributes.

''We find that social network connections contain valuable information about brand image,'' said Culotta. ''Traditionally, marketing researchers rely on customer surveys to gain detailed insights about how brands are perceived-but surveys can be costly and time-consuming.''

Cutler noted that ''Other recent approaches that involve analyzing the text of user-generated comments around the brand are limited as well because the majority of social media users don't actively write about brands-and even fewer will write about a brand's relevance to a particular attribute. Focusing on what users do, rather than what they say opens the door to new insights.''

Culotta added, ''Our automated approach allows nuanced attributes of brand image to be monitored at higher frequency and lower cost than was previously possible.''

The authors recognise that the motivations of individual consumers in following brands are likely varied and complex.

 But by aggregating the behavior of millions of brand fans, they average out such differences to distill out the underlying brand image.

The approach has limitations, however. The brand image scoring algorithm can be potentially gamed using Twitter bots or potential fake accounts created, bought and sold online to create illusions of support.