“Twitter is stupid, and Instagram is Twitter for people who can’t read.”
– Max (actress Kate Denning) from “2 Broke Girls.”
While I disagree with Max, there is an astute behavioral observation amidst her sarcasm: Photos are the easiest medium to create and consume. Why carefully craft a short, witty tweet about your favorite seasonal Starbucks drink when you can just snap a pic of your drink and throw a filter on it?
It’s not news that photos are becoming a bigger part of everybody’s lives. It’s even apparent at the Vatican. On Facebook, there’s now a viral “Then and Now” comparison of the crowds at Pope John Paul II’s funeral in 2005 and the inauguration of Pope Francis this year. The sea of illuminated mobile screens makes it clear that everyone is a photojournalist now – with tremendous influence over their own social audiences.
Let’s take a step back and review how the social landscape has made social imagery a priority:
• Pinterest is creating an image-based interest graph.
• Facebook makes a data-driven decision to focus its redesign on images.
• Twitter adds photo filters and makes tweets more visually appealing with cards in order to keep up with Instagram.
• Instagram will focus on big data.
Pretty pictures are great for users, but data is the real currency in the world of social networks. So where’s the data in photos? Hashtags are great, but is that data really reliable? I would argue that the connection of hashtag data to the content of photos is inconclusive because it’s a human who creates the correlation between the image and the word. Looking solely at words to generate photo intelligence reminds me of a scene from one of my favorite guilty pleasure movies, Zoolander, when male model Hansel (Owen Wilson) tries to look for files IN the computer.
The content of the photo is what is important to users, so that is the first place to look for valuable photo intelligence. Enter computer vision.
Computer vision is maturing as a science. Computer vision scientists are basically mathematicians who create algorithms that make sense of the 1’s and 0’s behind a digital image. In the same way that a CSI officer can identify a fingerprint at a crime scene based on a database of prints, a computer vision scientist can identify a product in a photo based on image indexing of the original product. Vision scientists have even made breakthroughs that allow for sub-pixel photo analysis in order to find objects in the background as well as the foreground of a photo.
“So what?” says the brand manager. “So my product is in that consumer’s ugly photo. I posted my pretty photo of my product on Pinterest and people repinned it.”
Yes, repins on Pinterest are great, but we are no longer in a push-marketing world. Penetrating social graphs and driving organic peer-to-peer influence in the wild are far more important than your marketing. Nielsen recently found that 92% of consumers around the world say they trust earned media, such as word-of-mouth or recommendations from friends and family, above all other forms of advertising. For every one of your repinned photos, there are potentially hundreds of thousands of more effective and organic photos in the marketplace.
Apply computer vision to a social network’s fire hose of photos and the intelligence output could answer some interesting questions:
* How do user-generated brand photos impact users’ social graphs?
* Do they influence friends’ opinions of brands or products?
* Do they influence friends’ willingness to interact with brands?
* Do they influence friends to purchase brands’ products?
* Are friends more likely to share a picture of a product if they see their friend’s photo of the same product?
* Are friends more likely to interact with brand’s social ads if they’ve just seen a friend’s product photo?
* Are friends more likely to interact with social ads if their friends influenced the creative? (Instagram beware.)
Images are the next realm of big data. While text mining isn’t obsolete, the sentiment of a brand mention in social media is inconclusive most of the time. A picture of a smiling consumer holding the same brand’s product is much more conclusive (and provable with computer vision). Since nobody wants to write (or read) a thousand words in a status post, extracting the equivalent information from social images is the next frontier of intelligence for social networks. Making this intelligence actionable can drive real results for brands.