Brand safety is the ultimate win-win for advertisers, agencies and publishers – so long as we ditch the blunt tools that don’t understand the context of content.
Nobody wants to see brands caught out by inadvertently advertising against embarrassing or unsafe content, such as hate speech or a story that is critical of the business or its industry.
It is no surprise that advertisers and their agencies are heavily invested in brand safety tools. However, the huge problem is that the type of technology many agencies and networks use, and so publishers have to accept, lets down both sides in digital marketing.
Inappropriate technologies misclassify content and as a result prevent publishers from monetising a considerable proportion of their inventory which is wrongly flagged as unsafe. This prevents advertisers from reaching targeted audiences who are engaged in quality content that fits in with their media plan.
Cursory glances are not enough
A very common problem lies in the use of keyword lists in an attempt to identify the context of the publisher’s page. Often a single word perceived as ‘negative’ can flag up a piece of content as unsafe. Many basic Brand Safety tools cross-check content at the page level against a ‘negative keyword list’ containing multiple words perceived as negative. To make matters worse other technologies may only look at the ‘URL’ without fetching the page and analysing the actual content behind the URL.
There are some very obvious downsides. A good example would be a tragic news story, such as the Manchester suicide bomb attack. Many brands may have understandably wanted to avoid advertising in an article talking about such a harrowing outrage. However, if they just take a negative keyword approach using specific keywords such as ‘Manchester’ or ‘Bomb’, the segment will block all content where those keywords are present. This could mean blocking an inspirational story about Man City FC or a recipe for making a bath bomb. This is a common error known as a ‘False-Positive’.
When keyword checkers go no further than a page’s URL, which can often be a mixture of words and letters, there have been even more bizarre cases of inappropriate content blocking. An example of this is a review of Christmas-make up gift sets which was incorrectly classified as ‘Weapons and Military’ due to the close proximity of the characters ‘A-K’ and ‘47’. Clearly a fashion client or any brand for that matter would be concerned about the quality of their vendor’s classification here, not to mention much needed and lost revenue to the publisher.
In both examples, a publisher has missed out on the chance to sell advertising against their content and an advertiser has been denied the opportunity to reach a targeted audience of people buying bath bombs in Manchester or searching for make-up gifts.
Reading like a human
What is needed, then, is a way of reading articles through advanced technology that better understands context. Rule-based Natural Language Processing (NLP) is now advanced enough to read an article and understand not just the words that are being used but their meaning, based on the complex relationships words have with each other, even enabling the ability to understand the sentiment and emotional qualities of the page. This approach reduces and in a lot of cases completely removes the issue of ‘False-Positives’ and ‘False-Negatives’ that penalise publishers unfairly but also leave advertisers missing out on quality audiences and content.
If one takes the example of the word ‘crime’ or ‘murder’ being used in an article, a keyword engine will block it, however a ‘rule-based’ NLP approach will distinguish between the content being a hard news story of a gruesome crime or if it’s with reference to a review of the hit television show ‘Line Of Duty’. Clearly many advertisers would like to keep their distance from a current murder report whereas they may pay a premium to appear next to Entertainment content such as ‘Line Of Duty’. NLP looks beyond the URL and processes all of the content on the page, including the grammar and syntactical elements which give words meaning and, wait for it….. Context!
Semantic contextual intelligence for better monetisation
The end result of false positives thrown up by keyword checkers is that publishers are missing out on a huge amount of revenue by not being able to fully monetise a proportion of their inventory. Lost revenue varies greatly but, in our experience, a double-digit percentage of many publishers’ entire output often struggles to receive advertising because it is wrongly blocked.
So, publishers have a very clear incentive to move the digital advertising industry away from keywords to a more sophisticated technology that better understands the nuances of the written word.
The trouble has been that, historically, agencies and networks have invested, on the demand side of the business (at DSP level), in keyword and URL checking tools. That means publishers are finding it hard to move the industry towards more sophisticated technology which would solve many of the problems raised by the industry today. With so much content to advertise against, advertisers may be forgiven for not worrying too much about some missed opportunities.
Offering advertisers better targeting
To gain traction, then, publishers need to use NLP tech to not only better understand their content, and stop it being flagged up wrongly as unsafe, but to also offer advertisers the additional benefit of segmenting audiences by interest based on a precise understanding of their reading and viewing habits.
It is no secret advertisers are willing to pay for targeted audiences on premium content pages. They want to reach the right people in the right context at the right time. Publishers may never match the sheer scale of Google or Facebook’s vast data reserves, but they can build a deep, rich understanding of the people on their site. Detailed profiles of interest groups (advanced interest profiles) can be put together so advertisers can buy premium audiences against premium content.
The elephant in the room here is that this is first party data. With GDPR now a year old and, separately, Google joining Apple in cracking down on third party cookies, the market is pivoting to recognising the immense power of first party data.
The immensely valuable first party data publishers hold here not only helps them dramatically boost digital revenues, it also helps advertisers in their mission to reach targeted audiences on quality sites. By working with sophisticated technologies such as Natural Language Processing advertisers and publishers can solve their Brand Safety woes but importantly leverage this deep contextual understanding to enrich their first party data.
Truly a win-win for all.
by Nick Welch, VP business development, UK and Northern Europe at ADmantX