top of page
Bob Hoffman

ALARMING RESULTS OF "ACCIDENTAL" RESEARCH


Among the frustrations of digital advertising, I think it’s fair to say that near the top is not knowing who and what to believe. While online advertising technology was supposed to provide us with near perfect data on who we were reaching, where we were reaching them, and what it was costing, twenty-five years later we find that the extent of uncertainty about what is happening with our advertising money online is staggering.

Between viewability issues, ad fraud, the “adtech tax,” “banner blindness,” agency and middleman fees and commissions, and “unknown deltas,” the extent to which a programmatic ad dollar buys actual advertising is highly uncertain.

A good example of the credibility gap is in ad fraud. Everybody agrees that there is some degree of fraud in the buying and selling of digital advertising, but the extent of fraud is the subject of wildly divergent estimates. Industry bodies and some fraud detection solution providers claim that fraud has been significantly reduced and now affects only about 1% of online ad dollars. Some independent researchers say that ad fraud is at least twenty times larger and steals over $80 billion annually from advertisers. How are we to know who’s right?

Ironically, non-digital (or, if you prefer, off-line) advertising is often able to provide more certainty about what we are paying, who we are paying it to, and what we are buying. If you buy an ad in The New York Times you can open a copy of The Times and see your ad there. If you buy a spot on Monday Night Football, you can watch the game and see your ad. But if you do a programmatic ad buy online it is virtually impossible to ascertain what you have bought or where it has run. According to the ISBA (Incorporate Society of British Advertisers) the average programmatic ad buy by a “sophisticated” advertiser distributes ads across 40,000 different websites.

There is no way for a media analyst to possibly get ”under the hood” of 40,000 different websites to determine what actually happened. Understanding the technological intricacies needed to do forensic analyses of computer-based advertising buying and selling is not a job for amateurs. In order to understand what is going on you probably need to read and interpret code, and be either a computer scientist or a software engineer. And even then many platforms will not provide file-level data, where the truth about your ad buy resides.

Consequently there is no way for any company’s management - no matter how sophisticated - to have first-hand knowledge of what is actually going on with all their digital ad spending. Instead, they must rely on reports they get from the systems used to sell the ads (often SSPs), the systems used to buy the ads (often DSPs), the people at agencies who gather the information and consolidate it, and the people in the marketing departments who evaluate and vet the reporting.

In other words, because of the arcane nature of the online advertising ecosystem, there is a chain of reporting that is the only way for businesses to understand how their digital advertising budgets are actually being spent.

The exasperating part of this is that there is also no way for the recipients of the reports to verify the accuracy of the reports themselves. So not only are the actions of the programmatic advertising system opaque, the reports that presumably validate the actions are also opaque.

But sometimes, just by chance, lightening strikes and it becomes clear to everyone what is actually going on. Such a thing happened recently. An “accidental” research project occurred which unambiguously demonstrated how much we can or can’t trust the competence of the experts who report to us and are charged with informing and protecting us from the opaque activities of the programmatic ad system.

In the clearest possible terms, we non-computer scientists, non-software engineers, and non-adtech experts got an explicit look at the inner workings of the people we rely on to inform us and protect us in this mysterious world.

Several weeks ago two researchers named Krzysztof Franaszek and Braedon Vickers, working at a company called Adalytics, stumbled upon an error. They discovered that Gannett Publishing had unintentionally been publishing online ads in the wrong places.

This chance discovery has lead to a uniquely revealing and disturbing set of revelations.

Gannett owns the USA Today website. It also owns hundreds of small community news sites. For nine months Gannett had been conducting billions of ad auctions for space on the USA Today site, but had accidentally been sending bid-winning ads to the other websites it owns. In other words, for nine months advertisers like Sears, Nike, Adidas, Ford, State Farm, Starbucks, Kia and Marriott had been paying for, and thinking, their ads ran on the USA Today site, but actually they may have run on the websites of the Ruidoso News, a biweekly newspaper in New Mexico, or the Lebanon Daily News in Lebanon, PA.

The question is this: How can billions of ads have run in the wrong places without...   

   ...a single brand noticing that their ads weren't where they were supposed to be?

   ...a single agency knowing what they were buying?

   ...a single fraud detection company, or media auditing firm unearthing the fact that billions of ads went to the wrong places?

According to The Wall Street Journal, at least fifteen different adtech companies that were part of the chain of buying, selling, and verification for Gannett had enough information to see what was going on. Not a single one of these companies discovered or reported it. As far as can be told, not a single company even understood what they were looking at.

Companies who had the information at hand were some of the biggest in the adtech field -- including Integral Ad Science, Double Verify, and Oracle's MOAT. Of particular interest is Integral Ad Science. Not only do these "scientists" specialize in reporting fraud, according to the Journal..."Gannett pays Integral Ad Science for insights on its traffic and metrics related to its advertising... Data gathered by the researchers and reviewed by the Journal showed that Integral Ad Science received information revealing the Gannett discrepancy thousands of times. Integral Ad Science didn’t inform Gannett of the phenomenon... and didn’t inform its advertiser clients, according to media buyers."


Several adtech companies that represent sellers on Gannett's websites including Pubmatic, TripleLift, and Criteo had enough information to know what was going on. Several companies that represent buyers on Gannett's sites, including Google's Display & Video 360, Publicis' Conversant, The Trade Desk, and Media Math also should have known. According to the Journal, "Each of them had enough information to raise concerns about the publisher’s auctions…”

Billions of ads from reputable companies ran in the wrong places and all along the adtech chain of responsibility companies that had the information necessary to see what was going on didn’t; not a single brand noticed that their ads were not where they were supposed to be; not a single media buyer noticed that their ads were misplaced. And for nine months we can only assume that these “sophisticated” advertisers were receiving fictitious reports about the nature of their programmatic buy.

If these leading adtech companies don't have the competence to discover billions of honest mistakes from a company who is not trying to deceive them, what level of confidence can we have in their ability to identify the work of fraudsters who are trying to deceive them? I think we know the answer.

Every once in a while serendipity strikes and shows us in unambiguous terms exactly what is going on. You could not design a clearer, more honest test of the competence and credibility of the programmatic advertising ecosystem.

We now have alarming and incontrovertible evidence - due to an accident - that we cannot trust the adtech ecosystem, and most vividly, we cannot trust the information we get from the people we pay to protect us from the uncertainties of the programmatic ad system.


[Next] [Home]

Comments


bottom of page