Visual Intelligence

Richard Lee, CEO of Netra, described during his Pecha Kucha talk that 2017 will be the year of visual intelligence. His view is based on a couple key factors. First, consumers are now taking & sharing more images than ever on social media. Simultaneously, camera equipped phones are now everywhere while digital storage costs continue to decrease significantly. Lastly, the social platforms themselves are becoming increasingly visual – Snap Inc. now refers to itself as a camera company and Facebook is not far behind in moving to this viewpoint.

If 2016 was the year of “voice” due to the increasing maturity of the technology and a larger variety of solid use cases, 2017 will be the year of “visual” due to image recognition technology improvements with the biggest and smallest technology companies now focusing on this space.

But if the technology has improved so much, why has adoption been so slow? Richard answered this referencing a quote from Steve Jobs that “people don’t know what they want until you show it to them.” The potential for image is huge: he noted 100% of all photos have context and if you can combine this context with the people behind them and with brands, you could start to target based off of these attributes.

Why is this so important, especially as it relates to targeting & product development? As Richard mentioned, “the photo feed is now the window to your soul – what you take pictures of shows what’s important to you.”

A Picture is Worth A 1,000 Words


David Rose from Ditto Labs and Sharad Verma from Piqora discuss the challenges and opportunities in visual media.

In Mary Meeker’s recent Internet Trends Report, she states there are 1.8 billion images shared daily across Flickr, Snapchat, Instagram, Facebook, and WhatsApp. This doesn’t contain any of the images shared on highly visual networks such as Pinterest and Tumblr or mixed media networks such as Twitter. The potential to analyze images and derive insights is huge; if a picture is worth 1,000 words then the potential value in analyzing images is at least as large as the value in analyzing text.

The conversation led with a discussion on the features of the networks that facilitate the creation and sharing of visual content. David and Sharad briefly debated whether mixed media networks would be able to harness the ease of consumption and emotional response that the more visual focused networks have used to catalyze their growth. Their conclusion was it is too soon to tell. They also discussed the motives of the users of the more visual networks citing discovery and self-selected feeds as reasons why people opt to use highly visual social networks.

David shared some stats around image analysis such as 130 million images a day are shared on Tumblr and of those images 28% contain text within the image that can be extrapolated. This began the discussion on types of analysis that are possible with images. David mentioned that 3-4% of those Tumblr images referenced were selfies and that there is a “smile score” where it is possible to quantify the emotion of the person in the image. They also shared some statistics around the cross-posting of images on various social networks such as 20% of Tumblr images get posted on Pinterest within two weeks of their initial post, and 40% of Tumblr images live on Pinterest. This indicates two types of users according to Sharad, people who cross-post strategically and optimize per network, and people who simply cross-post their content to as many networks as possible.

David shared a real-time feed of images from Tumblr, Instagram and Twitter which were related to clothing. He said that they can take feeds which contain filters on brand names, expressions, content and more and sell them to brands. According to David, approximately 85% of user-generated images on social networks that are relevant to a specific brand cannot be identified with text or hashtags, the image itself must be analyzed. Being able to do this allows brands to use social analytics to improve their consumer research, audience discovery, advertising, and understand the ROI of visual social posts.

E-commerce and the marketing funnel were strong themes throughout the discussion. Sharad made it clear that images posted on Pinterest are toward the bottom of the marketing funnel. They can link directly to a product page and can be analyzed for ROI easily, but measuring ROI beyond that direct response could be very valuable. Images on Instagram and other networks he considered more top of funnel and measuring their impact on ROI, along with things like category or price point, were very insightful in his experience. David mentioned the ability to analyze images from users on Twitter and find people with an affinity for a brand, or a competitor’s brand, and then reach out to them with tailored audiences could represent a unique marketing opportunity from visual social data.

Both agreed that social media analytics dashboards needed to be more visual and include more images to be more effective.