There’s no doubt that Instagram has fast become the shiny new kid on the advertising block and eMarketer’s latest forecast predicts that the image-centric social media platform is projected to rake in $595m in worldwide mobile ad revenues this year, before soaring to a whopping $2.81bn by 2017.
That’s pretty impressive growth considering Instagram only introduced its ad offering a year and a half ago. And with a raft of advertising products already launched and new features set to roll out over the next few months, it’s an exciting proposition for brands; particularly when you consider it has 300 million monthly active users and accounts for a large chunk of the 1.8 billion photos uploaded every day across social media.
A large part of Instagram’s success comes down to keeping advertisers on side by taking considerable steps to ensure that inappropriate image or video content doesn’t slip through the cracks. Last year, the Facebook-owned company inked separate deals with media holding companies Omnicom and Publicis Groupe for innovative partnerships to help build the infrastructure and tools to control the advertising process on the site.
Other visual sites should take a leaf out of Instagram’s book and step up their game by ensuring they have the necessary tools in place by giving brands visual protection against appearing alongside or in amongst ‘dodgy’ content – which could be something sexually suggestive, inappropriate or even violent as well as allow brands to ‘visually target’ ads alongside visual content.
Typically, the pictures that users post often come with little or no text description or metadata attached to them so advertisers relying on text-tagging will find it far more difficult to classify this content. What they must do is harvest visual data from the images themselves to get a steer on what each consumer is interested in, based on convolution networks – a type of deep learning architecture consisting of layers which can evaluate what’s in an image. Image recognition has the potential to determine sex, age, sentiment as well as the foreground and background nature / subject of a photo – that’s a lot of insight!
As the web becomes more and more visual, it’s now not sufficient to simply use contextual data to validate a picture based on the text around it. And the problem is that a lot of verification tools use data held of the user to moderate content rather than the actual content being viewed. This will always then be difficult to safely validate the images that are being uploaded.
Not only does image recognition allow brands to target advertising based on visual aspects of the page but will enable photo-centric sites to learn more about what their users share and consume. Such insights allow for smarter product development and business insight, as well as giving these visual sites the chance to compete for advertising dollars in a programmatic environment.
Accurately classifying content and introducing an automated solution on upload takes away the threat of an unsafe image being uploaded before being spotted by a moderator. Only with the necessary tools in place will enable Instagram’s peers to be more competitive and keep up in this visually dynamic online world, without getting left behind
By Adrian Moxley, Co-Founder and Chief Visionary Officer at WeSEE.
Originally posted Digital Marketing Magazine 31 July 2015