Less than three weeks after the Charlottesville protests (and barely six months after the Independent’s study on YouTube’s epic ad misplacements) a friend of mine posted an item from a fake news site to his Facebook page. The site, Mr. Conservative, included a photo of the creator, sandwiched between a Confederate flag and an ad for Walmart groceries. And there are plenty of other images to make a brand manager cringe, including people smoking from bongs. I’m pretty sure this site is a parody; but I’m dead certain Walmart wouldn’t want to be associated with it.
You wouldn’t be wrong if you feel like these brand safety gaffes aren’t slowing down. Just Google “misplaced ads” and you’ll see an abundance of listicles documenting the most recent worst offenders.
Why, when we have so many automated brand-safety tools at our disposal, is the industry still plagued by this issue? For my team, this isn’t a rhetorical question; we wanted to systematically find the source of these troubles.
Earlier this summer, my team randomly selected 1000 page URLs from premium general news sites. We then carefully reviewed the content on each site, and labeled them brand safe or brand unsafe (e.g. legitimate news stories that brands don’t want to be associated with, such as those about murders, drug rings, and so on). This manual effort ensured we had a clean dataset for testing purposes.
Next, we ran those 1000 URLs through four of the leading third-party brand-safety verification tools to see how they compared to the human-verified dataset. The results weren’t pretty. The top-performing vendor delivered a 71% match rate, the others matched at 53%, 29% and 26%. Conversely, the best of the bunch got it wrong 29% of the time, the worst 74% of the time.
Put another way, for three of the industry’s leading brand-safety verification vendors, a toss of a coin is likely to yield equal or even better results when assessing whether a page is suitable for a premium brand’s ad. In light of this, no one should be surprised that brand safety is still an issue! (Or why a brand manager would be quick to pull media spend from a marketing execution partner.)
What We Learned
I can’t tell you authoritatively why these results are so abysmal because how these third-party verification systems work are tightly kept trade secrets. But here are things we know, and things we don’t know.
One of the things we do know is that all of the providers rely on text analysis alone, which clearly isn’t enough. Images that are disturbing -- flooded streets of Houston -- will certainly appear in every legitimate news site, and must be screened for separately. (Other marketingblogs, such as the Houston SPCA or the American Red Cross will opt for these opportunities.)
One of the things we don’t know is whether the third-party vendors screen each and every impression, or if they rely on sampling, as many do when calculating viewability rates. But I will say this, sampling has inherent limitations, which means it’s critical that these companies process every single impression.
Where Do We Go From Here?
I think we’re at an interesting inflection point. Our industry has witnessed tremendous progress in terms of machine learning, automaton, and artificial intelligence. These technologies are deployed in myriad ways to ensure the right creative is displayed to the right customer, in the right channel, at the right time. Why can’t these technologies be leveraged to push for brand safety?
Equally important, the industry as a whole needs to rise to the occasion. With brand safety, I sense an attitude of, ‘as long as I’m hitting the checkbox of one of these third-party verification systems, I’m okay.’ There is no real sheriff to ensure these systems are at a level that’s acceptable to brands. In fact, if the Independent hadn’t released its YouTube study, I’m not confident anyone inside our industry would have questioned the status quote.
But outside our industry, that status quo is being questioned. Citizen activist groups, like Sleeping Giants, are actively monitoring offensive sites for the purpose of “outing” brands that are unwittingly funding hate through advertising. Their efforts are creating a perilous environment for marketingblogs, one in which a single ad misplacement can have negative consequences. And I don’t blame them for taking such actions, this has got to stop.
As an industry, we need to fix this problem the same way we’ve addressed every other challenge we’ve faced, namely taking full responsibility for it, and leveraging the best available technology to address it.