Image Recognition and the Future of Digital Analytics

The days of text-centric social feeds are officially long gone. A whopping 1.8 billion images are uploaded to the Internet daily and of those, 350 million are shared on Facebook. Instagram recently surpassed 500 million active users, and Snapchat now has more active users than Twitter. The content that flows into our social feeds is more heavily optimized than ever to deliver more of what people want—less text and more visuals.

Brands have adapted their social content strategies accordingly by delivering more visually immersive experiences. And while we’re seeing significant shifts in branded content, this influx of visual content has yet to herald a commensurate change in social analytics. Accordingly, few gains have been made to measure and derive insights from the contents of images or video. Social listening has historically focused on the challenges of text-based analysis–specifically, the challenge of determining the context and meaning behind posts. But as social media habits evolve, it’s clear that deriving insights from pictures is an increasingly important aspect of understanding consumers. That’s where image recognition comes into play.

Brands have adapted their social content strategies accordingly by delivering more visually immersive experiences.

Simply put, image recognition is the process of translating images to data. Photos and images can reveal a wealth of data points–demographics, purchases, personalities, and behaviors (just to name a few). Through next generation image recognition, a mere selfie may reveal a person’s gender, approximate age, location disposition, and even the clothing brands that the person is wearing. As text-centric media takes a backseat to image and video, the opportunity to understand the contents of these formats grows. These insights represent a veritable treasure trove of actionable data for brands.

Tools that analyze image and video-based content are still in development, but increased investment in research is already impacting commercial products and how they’re advertised. One example is brand logo recognition–scanning images for brand logos, and flagging them with the corresponding brand names. This tool is especially powerful considering that 80% of photos shared online depict a brand logo but don’t explicitly call out the brand’s name. This fact points to a sizable opportunity for companies to measure and understand the impact of these formerly inaccessible data points.

Photos and images can reveal a wealth of data points–demographics, purchases, personalities, and behaviors (just to name a few).

As an example of how this applies to brands, Kelton’s Digital Analytics team took a look at the scores of backyard BBQ photos that flooded public forums, blogs, and social feeds over the recent 4^th of July holiday. We experimented to see which of two quintessentially American beverage brands–Coca-Cola and Budweiser–netted more published images of patriotically-themed bottles and cans (as well as other forms of branding) on social media.

In the end, Coca-Cola branding was twice as prominent as Budweiser’s. We found that Coke bottles and cans popped up in more diverse settings such as public parks and inside motor vehicles, whereas Budweiser was predominantly found in bars and house parties. Coke also aroused greater sentiment around the theme of Americana, as many consumers photographed vintage Coca-Cola gear and opted for bottles over cans. This might explain why Coke captured a significantly greater share of social mentions than Budweiser.

This example illustrates several ways that brands can leverage image recognition technology to build actionable insights:

• Ethnographic data – Identify where, when and how often brands are showing up in people’s lives.
• Updated brand health analysis – We now have a more comprehensive point of view of brands’ online footprint.
• Sponsorship and Branding ROI – Extend the value of branding and sponsorships shared via online news, blogs and social media through a multiplier effect.
• Influencer identification – Find authentic brand advocates who consume and spotlight your merchandise.
• Misuse use of brand iconography – Surface content that depicts improper usage of brand’s logo or other creative assets.

In today’s ever-shifting social media landscape, it’s never been more important for brands and their partners to stay aware of the new and emerging capabilities that can help better understand consumers’ behavior online. Image recognition is just the beginning. From AI startups to instant objection recognition devices, the mobilization and fusion of research, tech, and capital is quickly reshaping the way we think about analytics. These new tools will add even more contextual understanding to sentiment on social platforms, empowering brands to understand consumers like never before.

Liam Freaney also contributed to this blog post.