The growth of the automatic targeting of online and social advertising is hitting a language barrier. But don’t panic, WeSEE’s Adrian Moxley has the answer.
It’s not only in the West where social media advertising revenues are growing rapidly. In the Asia-Pacific region they’re rocketing. In fact, China is fast catching up with the market leader, the US, with eMarketer predicting revenues of $3.4 billion for 2015, almost doubling to $6.1 billion by 2017. To add some context, this represents around 12.5% of all digital ad spend in the country.
But China is not alone in its hunger for social advertising. The Asia-Pacific region overall is performing well, with eMarketer forecasting 2015 revenues to reach $7.4 billion, a huge 43% jump from 2014.
This is great news for publishers and social media channels, as well as for advertisers who can tap into this increasingly socially active audience. It also has the potential to push the already fast-growing programmatic sector through the roof. But there’s just one problem, and it’s quite a significant one – language.
The simple fact is that the majority of programmatic advertisers use contextual – or at least text-based – data for targeting ads. And very few data partners offer global reach in terms of languages and dialects. This can cause complexities and create barriers for brands looking to target global online and social audiences with their advertising.
So what’s the answer?
Well we all know that a picture paints a thousand words. Or did you think that was just a stupid saying?
But seriously, most online and social content these days is accompanied by at least one relevant image. And this image acts as a signpost to what the content is about, no matter what language the text is in, which of course holds the key to effective programmatic targeting.
For example, if a Japanese language website carried an image of a baseball game, it would be clear what the subject matter of the article was without the need to analyse the text. Likewise, if a Khazakstani site displayed an image of woman in a black dress walking on a red carpet, the content could be easily classified as entertainment and/or fashion. Essentially, you can determine the nature of the page from its visual content in most cases, even content that contains negative or innaoproriate subject matter.
Recent research actually shows that if you take the image into account you can tell the story of what the content is about on two-thirds of all web pages. Therefore, using image recognition and classification technology could be the answer to unblocking access to the global web for advertisers and brands.
But images can actually provide more information than text. The same research also suggests that for one-third of web pages, the nature of a page can radically change because of the image. For example, a travel blog on Menorca would generally only be classified contextually as travel and perhaps Spain. However, it might be relevant to other advertising categories, which can only be identified through image recognition.
For instance, if the travel blog contains an image of a group of women sat around a 4×4 sipping a beer, suddenly the page could be relevant to food, drink, automotive or fashion, yet this is unlikely to ever be picked up through text classification. So, as well as opening up the global web to advertisers, using imagery can help grow the potential online and social ad inventory.
Get the picture?
By Adrian Moxley
Co-founder and Chief Visionary Officer