In one instance, Facebook users clicking to read a story about Michelle Obama’s encounter with a 10-year-old girl whose father was jobless were then offered a "related" story alleging a Secret Service officer had found the president and his wife having “S*X in Oval Office,” and another that said “Barack has lost all control of Michelle” and was considering divorce. Facebook's explanation: algorithms that decided the tabloid content was a good match for readers who track Michelle Obama.
In the second case, an image of a nude American model on a porn site ran next to an Australian government ad asking the public if they would "prefer asset sales or service cuts?" The explanation? Amusingly, a contextual targeting mishap, with the technology not distinguishing between the term “cuts” within the ad copy that referred to government services, and the “cuts” on the Web page, which referred to circumcision.
These were not the first (or the last) instances of machines failing man, nor are they even the most egregious examples. Whenever the guilty party is called to task for errant online ad activity, they immediately point to fault lines in their automated calculations that failed to make the distinctions they promise in their promotions, inevitably claiming that this was a "rare exception" and that everything works fine most of the time. The problem is that those "rare" exceptions tend to tarnish not just the brands who trusted the technology in the first place, but the entire online ad industry
Meanwhile, Magna Global forecasts the global programmatic ad market will grow from $12 billion in 2013 to more than $32 billion in 2017. Part of the alleged attraction of programmatic ad buying and selling (aside from monetizing otherwise hard-to-move inventory, consolidating customer interactions across multiple channels into a single dashboard, blah, blah, blah) is that it eliminates the potential for human error, assuring brand safety for all. Really?
It seems not a day goes by without a report on some failed aspect of online video advertising, from never-seen below-the-fold auto plays that are counted as served impressions to volumes of bot traffic that count as valid human views. Just yesterday The Wall Street Journal reported on ads for major brands including Kraft, Toyota, Target and Honda appearing alongside pirated streams of popular TV shows including “Game of Thrones” and “Mad Men.” The sites were impossible to reach and tended to disappear, then reappear when pressure is applied. Meanwhile, somebody gets paid for placing the brand ads. My bet: automated all the way.
While it could well be argued that these are issues with the scores of individual companies participating in programmatic platforms and not the buying and selling platforms themselves, I suspect that anywhere algorithms are asked to make distinctions that could realistically be made only by human judgment, errors will be made. And at least of part of those errors will be intentional fraud, if only through neglect.
After all, you have an online ad ecosystem that rewards fraud because lots of folks get paid for ads running in the wrong place or unseen by real eyeballs. Meanwhile, brands -- often unaware of such activity -- don't hold either their agencies or ad tech companies more accountable. Why? Because the spend is so insignificant compared to their TV budgets? Or because they are assured that everything works "as promised" BECAUSE of the automated processes in place?
I am not naive. I understand that automation has greatly benefited the online ad world and attracted lots of dollars that might have otherwise gone elsewhere. But unless the industry cleans up its act, sooner or later a burned brand will announce they are no longer going to participate in programmatic processes -- and others will follow.
In an amusing side note, some of the greatest benefactors of fraudulent traffic and unseen ads -- Google, Facebook, AOL and Twitter -- yesterday formed TrustInAds.org, a new organization aimed at "identifying scam trends and protecting consumers from malicious online ads."
That's nice, but it doesn't address the fundamental automation issues that allow both fake traffic and brand safety problems.