�鶹ӰԺ

Historically, social media listening has relied on text-centric posts to provide insights into consumer behavior and preferences, but that is no longer the case.

Instagram’s popularity has soared in recent years, and its consumer-created image-dominant posts are just one example of how photos and other visual imagery are replacing text-centric posts on social media. 

Images created and shared by consumers on social media offer companies a new way to study consumers’ perceptions and attitudes toward brands in real time. In a clear case of “a picture is worth a thousand words,” image-posts convey the impressions and feelings consumers associate with brands without them saying a word. 

Leeds School of Business Assistant Professor Liu Liu and colleagues Daria Dzyabura of the New Economics School in Russia and Natalie Mizik of the University of Washington realized that these consumer-created images would provide companies a rich source of information about how their brand is perceived and experienced.

In a first-of-its-kind study, Liu and her colleagues encourage companies to sit up and take notice of this practice—or risk missing out on valuable, targeted information shared freely by customers about their brands.   

Prada or Eddie Bauer—who’s more “glam”?

Liu’s team introduced a visual listening approach to monitor consumer generated visual content on social media. To do so, they developed BrandImageNet, a tool for firms to gain a deeper understanding of their users’ brand-related “conversations” online, and ultimately enhance their brand management. 
 
The BrandImageNet model maps the images consumers create and share to specific perceptual attributes, such as “glamorous,” “rugged,” etc. For example, is an image portraying Prada as a glamorous brand, and if so, is it portrayed as more glamorous than Eddie Bauer? Firms may then measure how their brand is represented online in relation to those attributes. 
  
“Given that images are on their way to surpassing text as the medium of choice for online conversations, monitoring visual content is important to get a more complete understanding of online conversations involving brands.”

Deep insights from AI  

BrandImageNet is part of a new world of artificial intelligence being applied to research, including social filtering, that imitates the human brain’s functions for decision making.
 
The model itself is based in deep learning, a subset of machine learning, that is used to identify and analyze visual imagery by simulating the activity of the brain and central nervous system.  
 
Liu’s team applied the BrandImageNet model to both company- and consumer-created images on Instagram for 56 total apparel and beverage brands—the categories most frequently posted by consumers. The model was trained to identify and analyze the brand attributes in the images and what they represent to consumers. 
 
The metrics provided by BrandImageNet were apparent: Consumers’ representation of their experiences with brands on social media clearly reflects their brand perceptions.
 

That information is gold for companies and their marketing and advertising teams. 
 
Liu and her colleagues’ study demonstrates that BrandImageNet can offer companies an array of benefits, including a better understanding of who their customers are, how their brand is perceived, and gaps in their own positioning strategies to fine tune their messaging. Using the BrandImageNet model, firms can keep tabs on how their brand is portrayed in images on social media and compare it to competitors as well as their own brand positioning. 
 
The takeaway for companies? Go forth and better target your ads; the insights are in the images.