At least 15 advertising-technology companies had sufficient information to detect that publisher Gannett Co. GCI 3.44% provided inaccurate data to advertisers for more than nine months, but the firms failed to connect the dots and alert their clients, according to researchers who studied the incident.
Gannett, owner of USA Today and news outlets in 46 U.S. states, said it unintentionally provided the wrong information in online ad auctions, due to an error that affected its systems from late May 2021 to early March of this year.
The discrepancies meant that advertisers thought they were buying an ad on one Gannett site, but actually purchased space on another, the researchers said previously, estimating that billions of ad auctions were affected. The Wall Street Journal first reported on the Gannett error in March.
Many leading ad-tech companies had enough data to see the discrepancy, according to the team that studied it—independent ad industry researcher Braedon Vickers and Krzysztof Franaszek, founder of Adalytics, a firm that helps brands analyze where their ads appear.
The firms that were in position to sound the alarm include companies that specialize in putting ad space up for sale in online auctions, companies that assist advertisers in purchasing ads, and firms that specialize in detecting ad fraud, the researchers said. A selling point for ad-tech firms is their ability to validate that traffic for sale in online marketplaces is legitimate.
Some publishing and ad executives said the situation at Gannett raised concerns about whether the industry is missing other substantial discrepancies or intentional, fraudulent behavior. A practice known as “domain spoofing,” in which low-quality websites pose as brand-name sites in ad auctions, has become a significant problem for advertisers.
“There are actors who mischaracterize their content for their own financial gain. You have to think—man, the system’s not catching that either,” said David Chavern, chief executive of the News Media Alliance, a trade association that represents news publishers. “If they aren’t catching an honest mistake, then they certainly aren’t catching that either.”
In automated or “programmatic” online advertising, publishers put their ad space into an auction with important information for potential bidders, including details about the site where an ad will appear.
Companies that specialize in detecting ad fraud, including Integral Ad Science Inc., DoubleVerify Inc., Pixalate Inc., Oracle Corp.’s Moat and Human Security Inc., received enough data from Gannett’s auctions to spot the discrepancy and flag it, Messrs. Vickers and Franaszek said.
Gannett pays Integral Ad Science for insights on its traffic and metrics related to its advertising, according to a Gannett spokeswoman. Data gathered by the researchers and reviewed by the Journal showed that Integral Ad Science received information revealing the Gannett discrepancy thousands of times.
An Integral Ad Science representative said the company collects information needed to spot mix-ups such as the one in Gannett’s auctions, but she didn’t address Gannett’s situation.
Integral Ad Science didn’t inform Gannett of the phenomenon, the Gannett spokeswoman said, and didn’t inform its advertiser clients, according to media buyers.
“Mismatch alerts are something we are considering creating for publishers in the future,” the Integral Ad Science representative said.
DoubleVerify, another firm that validates whether ads appear where they should, contacted clients after the Journal first reported on Gannett’s error, saying its systems had detected the issue, according to people familiar with the matter.
SHARE YOUR THOUGHTS
What’s the best way to prevent fraud in the digital ad business? Join the conversation below.
A DoubleVerify spokesman said the company’s analytics product caught the discrepancy. The DoubleVerify spokesman had no comment on whether the company notified Gannett or its clients. The spokeswoman for Gannett said the publisher wasn’t notified.
A Human Security representative said the company’s monitoring system picked up on the discrepancy but didn’t flag it as fraudulent, so Human didn’t notify clients. Moat declined to comment.
Another firm, Pixalate, said its software picked up on Gannett’s issue while it was ongoing. Pixalate made this information available in its customer dashboard, according to a Pixalate spokesman. Pixalate said it marked websites for USA Today and other Gannett publications as high-risk domains and that advertiser clients could set up custom alerts to receive such information. Pixalate didn’t flag the issue to clients by default, according to people familiar with the situation.
Gannett worked with several ad-tech companies to sell its ad space. Each of them had enough information to raise concerns about the publisher’s auctions, the researchers said. Several companies received both the erroneous auction data and the correct auction data, including PubMatic Inc., TripleLift Inc., Magnite Inc., Criteo SA and Index Exchange Inc., according to the researchers.
A PubMatic spokeswoman declined to comment. A Magnite spokeswoman didn’t respond to a request for comment. An Index Exchange spokeswoman had no comment. A Criteo spokeswoman confirmed its partnership with Gannett but provided no further information.
A TripleLift spokeswoman said the company didn’t rely on the flawed auction data or pass it to ad buyers.
Advertisers bought ads on Gannett’s properties via ad-buying software provided by a variety of companies. At least five providers of this software received sufficient data to detect the Gannett discrepancy, according to the researchers: Alphabet Inc. -owned Google’s Display & Video 360, Adobe Inc., Publicis Groupe SA -owned Conversant, The Trade Desk Inc. and MediaMath Inc.
A MediaMath spokesman said the company didn’t rely on the flawed auction data.
Representatives for Google and Trade Desk declined to comment. Representatives for Adobe and Publicis didn’t respond to requests for comment.
Write to Patience Haggin at [email protected]
Copyright ©2022 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8
