Something is (terribly) wrong with Brand-Safety today
A couple of months ago, I came across a video campaign which caught my attention. It was an ad run by one the world’s leading brands, on one of the largest and most respected publishers in the US. The video itself was cheerful, smiley and ingratiating, and displayed the brand in all of its glorious positivity. There was however one small problem – It was running on a video article about the rape and murder of an eight-year-old girl. Seeing this beloved and joyful brand on display alongside such horrific content, completely negated the message they were trying to convey, and turned a positive experience into a nightmare.
Now, as it turns out, this advertiser was using the services of a prominent ad-verification platform, and yet their ads were being served repeatedly on extremely negative content. In fact, according to recent surveys – 3 out every 4 Fortune 500 brands are reporting repeated exposure alongside brand damaging content . No wonder Brand-Safety has become the number one priority for digital marketers, according to a recent report by eMarketer.
“It’s the tech, stupid” – Why first-generation solutions can’t deliver real Brand-Safety
Have you ever stopped to consider just how hard it is to enforce brand-safety? To protect advertisers, platforms need to be live on the page, and able to analyze all of the content and meta-data in their entirety, at incredible speed, and determine whether the content is brand-safe or not. Add to that the fact the even the most advanced computer algorithms have language capabilities that don’t even come close to those of a four-year-old child, and what you get is very poor content analysis capabilities.
So, do you want to understand how first-generation solutions go about enforcing Brand-Safety. Well, this is how it works – They upload a list of several thousand negative keywords (i.e. Murder, Terror, Sexual Assault), and then look for those keywords as they appear in content. If those keywords are found, then the content is flagged as negative. Sounds pretty straightforward, but ask yourselves – what happens when they come across an article containing a sentence like “Kevin Durant killed it on defense tonight”. With today’s solutions, what you get is a perfectly positive article about a basketball game, being flagged for containing one word which happens to have a negative interpretation. Now imagine that pretty much the entire suite of first-gen solutions enforce brand safety using this one simplistic practice, and you understand why advertisers aren’t getting much protection. Also, the analysis engines which are out there today are so dated, that they can’t even report in real-time, which means that they don’t actually prevent ads from being served to flagged content, but rather – provide the advertiser with a report, after the damage is done.
The result – Advertisers are getting damage reports instead of damage control
When you’re a leading global brand, you don’t want your ads appearing on horrific content, only to get a report from your brand-safety vendor telling you that you’ve been screwed over. Not only does this not protect the advertiser, but it creates more and more work on his part. Today, we are witnessing advertisers and agencies setup entire “Brand Safety” departments, with teams of analysts whose sole purpose is to read these damage reports, mine for insight, present the findings and optimize for future campaigns. All this manual labor going into an ineffective process which provides little value.
The rise of Autonomous Brand Safety, and how advertisers can take back control
What do advertisers really want? Do they want damage reports, or damage control? Do they want simplistic content analysis, or do they want high level efficiency? The answer is obvious, the question is – how do we achieve the latter? Next-gen Brand-Safety relies on mastering two technological components – Accuracy and speed. Accuracy requires us to evolve from simplistic keyword lists to advanced use of Natural Language Processing (NLP) modules. The ability to understand the true meaning of a word, in context, or to accurately label an article as positive or negative, requires training algorithms to better understand the content. The second component – speed, requires us to perform all of this complex analysis in under 30 milliseconds, enabling us to proactively prevent the ads from serving on negatively labels content. When you match sophisticated NLP with extremely fast and cost-efficient analysis, you get fully autonomous brand safety – A plug and play solution where brands can serve their ads without fear of negative placement, and without the extensive manual labor which is required to analyze futile damage reports.
To better understand the value of Autonomous Brand Safety, here’s what’s about to change for advertisers as they adopt this disruptive technology:
From simplistic tech to advanced NLP
The days of “dumb” keyword lists are over, as advertisers can expect deep content analysis and accurate content labeling. Some of the practices being put to work in Autonomous Brand Safety include Lexical Semantics (understating words in relation to context), advanced topic modeling (clustering words into larger groups and topics for the purpose of labeling), and audio content analysis (word segmentation, sentiment analysis). This will not only provide real contextual understanding, but also expand advertisers brand-safety efforts across new mediums such as audio and video.
From detection to prevention
Measurement, monitoring, and reporting don’t make a brand feel safe. In fact, Facebook have integrated 24 measurement platforms since 2008 and yet they’re still perceived as the least brand-safe platform according to marketers (GumGum Brand Safety Research 2018). Autonomous Brand-Safety allows for accurate analysis and rapid decision making, which means that Brand-Safety violations can be detected before the ad is served, thus preventing the advertisers from ever appearing in a negative environment. This is not “pre-bid” brand-safety, a term generally referring to the pre-bid filtering of inventory, but rather a real-time, per-impression analysis of every single impression and an on-the-spot decision. This means real accuracy, in real-time, and actual prevention.
From manual monitoring and optimization to complete automation
One of the greatest pain-points that have been left unaddressed in the age of Brand-Safety, is all the manual labor that goes in to monitoring, analyzing and optimizing. First-Gen solutions are bombarding agencies and advertisers with complex brand-safety reports, and they’re left to deal with the mess, while their ads continue to serve in compromised environments. Agencies are building entire teams of analysts to deal with the heaps of information, while the best they can hope to achieve is incremental improvement. Autonomous Brand Safety is about delivering a hands-off, plug-and-play, fully automated solution which does the work for us. A platform which detects and blocks violations in real time and in accordance with the brand’s own guidelines. This is not just about better results, and real-protection – It’s about saving time and energy on the advertiser’s side and making sure they don’t need to worry about these hazardous issues ever again.