A Spoonful of Visual Listening: Boosting Your Social Strategy

By: Dheepan Ramanan

March 31, 2016

social listening
social media
visual listening

“I see,” said the blind man.

Social media managers have long sung the praises of pictures to create more robust brand engagement. One study estimated that between 54% and 75% of Facebook brand posts are images. Pictures are innately eye catching, and the effects are most felt on Twitter. The average picture tweet garners 313% more engagement on average than text alone. Due to the potency of these posts, brands have carefully calibrated their internal strategies to eke out every last bit of value from this social media phenomenon. But this misses half the picture. Organic customer generated photos are much harder to monitor, tag, and analyze. Enter visual listening.


Visual Listening Gives Brands Clear Eyesight On Social Media


Using sophisticated deep learning neural networks, brands can spot their logos on social feeds everywhere without keyword mentions. This technique is referred to as visual listening. This form of social listening is critical: last year an average of 1.8 billion photos were posted per day on social networks. The unfiltered nature of these mentions provides huge value. Over 85% of images with brand logos do not feature a brand keyword.

Without visual listening, critical customer service issues may go undetected and marketing opportunities may be missed. In our study of General Mills, we uncovered insights focused around these two core use cases.

We analyzed six different General Mills brands: Betty Crocker, Cheerios, Nature Valley, Pillsbury, Totinos, and Yoplait. The above chart demonstrates that visual listening detects data that text-only listening misses.

Of the 6 brands analyzed, photo mentions comprised about 18% of all feedback.

30% of all Nature Valley feedback mentions came from images.

Photos are more organic, only 0.3% of photos contain #sp for sponsored tweets. compared to 2.2% of regular tweets.


A Bird’s Eye View On Twitter: General Mills Brand Analysis 


When analyzing text-only tweets compared to tweets containing images, the six brands differ in their share of voice.


  • There are more posts about Cheerios than any other brand.

  • Nature Valley has a larger share of voice from images than it does from plain text tweets.

  • Pillsbury has twice as many Twitter posts as image mentions, Totinos has a similar ratio of images to tweets.


Within Ear Shot: Listening to Customer Feedback on Twitter




tweets found through visual listeningGeneral Mills customers vocally expressed their thoughts about the brands we studied. These range from positive emotions for the brands themselves, to frustration in seeking customer service.

  • The Likes category, a topic that measures generic positive brand comments, is the most mentioned topic in text-based tweets.
  • Coupons/Vouchers are the most common Twitter theme.
  • Ingredients were the fourth most mentioned theme.
    • This was due to a GM recall of gluten-free cheerios that contained gluten.
    • The theme appeared with equal frequency in both tweets & image posts.
  • Box Tops for Education are more common in visual listening posts.
  • Packaging is far more mentioned in image posts.


Prescription Lenses Needed: Social Customer Service Oversights Without Visual Listening

Packaging is most mentioned by Pillsbury customers, with twice as many proportional mentions on this topic. A few of these comments concern defective packaging.  Although customers have learned to contact brands on Twitter, many may contact the non-service handle or fail to include a Twitter handle at all.

This comment (to the right) in particular shows how visual listening can pickup on a customer care issue that is not tagged.
A picture may say a thousand words, but only if a brand sees it. An average of 10% of brand mentions are images, the majority of which contain no identifying keywords. It’s clear from our analysis in order to obtain comprehensive information about customer feedback brands need to get serious about visual listening.





Dheepan Ramanan is a data scientist at Clarabridge. Follow him on Twitter @DheepanRamanan.