How Coca-Cola, Nissan and Kraft mine selfies for 'invisible' insight

Image recognition: brands can mine photo sites for insight
Image recognition: brands can mine photo sites for insight

Brands are slowly finding ways to make sense of image-based social networks such as Instagram and Tumblr, mining user photos for insight into how their products are used.

The rise of the 'visual economy' poses several challenges for marketers. One is understanding how consumers like to share post-production edits of their lives. Another challenge for brands is finding a way to participate in that activity with authenticity. And the final one is how to parse this visual data for insight.

David Rose, founder of image recognition startup Ditto Labs, thinks he has found a way to solve the last problem and give brands a way to "quantify" how consumers use their products.

Brands want to see how their product is actually being used in the wild, who’s using it and where - they want to understand usage occasion

Ditto’s technology scans photos for logos and recognisable patterns, such as Burberry’s famous tartan print or the Nike swoosh, detecting how consumers might be using products. According to Rose, Ditto’s system can recognise thousands of mainstream brand logos and patterns – even if a product is 80% obscured in the photo.

"Ditto finds those patterns even if they are around a can, chest, hat, skewed or partially included," Rose told Marketing.

The company recently struck an API deal with Tumblr, meaning it can trawl the site for user photos, spot products featured in images, and pass on the results to brands. Ditto also has access to Twitter, and partial access to Instagram. 

It is already working with one of Coca-Cola’s US agencies, Nissan and Michael Kors. Rival Curalate operates in the same space, scanning Pinterest and Tumblr on behalf of brands such as Gap and Urban Outfitters.

"Brands are using this for three reasons," Rose told Marketing. "One reason is in the way brands would do a focus group, they want to see how their product is actually being used in the wild, who’s using it, where - they want to understand usage occasion.

"The second is being able to look at a competitive set; how often is my product being used, and where versus my competitors."

He added: "Finally there's the relation matrix - we call it affinities. If we find an image of a Michael Kors bag, and you look at the user’s photos, what else are they interested in? Are they a foodie, are they into sports?"

How useful is image recognition for brands?

Not everyone is convinced that wholesale data gathering in this way is a useful long-term activity for brands. Forrester analyst James McQuivey said brands should focus on participating in social media over the long term, but observe it during big campaigns and other activity spikes.

He said: "I think Ditto will end up being like a type of focus group or ethnographic research. It can be very helpful at specific moments in a brand's lifecycle, but it's probably not necessary to track every day, all year long."

"Participating and interacting with social media users can be useful every day, especially for brands that need to learn how to get close to digital consumers for the first time," he added. "These two things can co-exist, but one is more likely to be a long-term behaviour than the other." 

Rose argues image scanning can go beyond insight and help brands engage with consumers directly.

For example, if a drinks brand such as Guinness wanted to target customers in Newcastle, Ditto can scan Twitter for location-based snaps of users drinking rival products, then hand over the data. That would enable Guinness to target those users directly through coupons or offers.

"It’s not just an insights platform, but a direct engagement platform," said Rose.

Creeping out users

Given user sensitivity around the handling of their personal data and photos, however, brands should choose "appropriate moments" for targeting.

"Brands have to determine how much they are a passion brand," said Rose. "They have to decide whether [an approach] is appropriate. There’s very rarely an issue with approaching someone wearing a Manchester United t-shirt in a photo and saying, 'hey, how about tickets to the game?'"

And it is not just users that might be miffed about how their photos are being used – brands might also make some unpleasant discoveries, according to McQuivey.

"Companies may not be pleased to find out how their products are used," he said. "It might be easier to pretend that all customers are ideal and their product experiences are ideal. Detecting their use of you on social media will quickly force you to expand your understanding of your customer."

Read next

Five behaviours marketers can learn from the new YouTube stars

How human-computer symbiosis is driving the future of big data

Subscribe today for just $116 a year

Get the very latest news and insight from Campaign with unrestricted access to , plus get exclusive discounts to Campaign events

Become a subscriber


Don’t miss your daily fix of breaking news, latest work, advice and commentary.

register free