No one said ad tech wasn’t complicated.
If a brand like Pepsi pays $10 to run an online ad with The New York Times, a rational expectation is that the media company gets to pocket the majority of revenue, with ad-tech vendors taking a small fee for processing the transaction.
But in some cases, media companies like the Times only get a couple of cents, with the middlemen pocketing the majority. Other times, premium publishers have to give 7% more of their revenue to an ad-tech vendor than a clickbait-heavy, made-for-advertising site would.
These are some of the findings in a recent study from ad-tech measurement firm Adalytics, which crowdsources advertising data based on a browser extension.
Using open-source techniques and some private data, Adalytics’ primary researcher, Krzysztof Franaszek, found that the revenue publishers receive from digital ad auctions is often highly unpredictable.
Of course, Adalytics’ research isn’t the first to shed light on the opacity of ad tech. A landmark 2020 ISBA and PwC report found that publishers on average received only 51% of ad spend and that 15% of digital advertising spend couldn’t be accounted for. Another study published Tuesday, from Franaszek and another independent researcher, Braedon Vickers, found that advertisers were misled about where their ads with publisher Gannett were placed.
Adalytics data matches media buyers’ bids and impressions, or publishers’ inventory, using a number of forensic methods. With a sample of 1 million pairs of impressions and corresponding bids from the fourth quarter of 2021 and the first quarter of this year, Adalytics found several ways ad-tech providers deprive publishers of revenue.
The majority of the publishers and ad-tech companies interviewed by Adweek said the findings rang true of their frustrations with some ad-tech vendors.
“It’s unclear to what extent these intervening technologies are useful,” said one publisher ad-tech executive, who wasn’t authorized to talk to the press. “That [ad-tech vendors] provide value for what they charge has always seemed unlikely to me.”
Expensive take rates
The study found that for half of ad impressions, middlemen’s fees range from 22% to 45%, with an average of 35%. At the extreme end of the scale, the publisher only received pennies on the dollar for a number of ad impressions in the study, with ad-tech vendors taking 80% to 98% of the transaction.
Tech providers can charge exorbitant take rates, sources told Adweek, though cumulative fees as high as 98% are uncommon. Franaszek said 2.73% of the data set had ad-tech take rates of 80% or higher, and only 0.07% were as high as 98%. Plus, there is great variability among publishers; for some, a higher percentage of their impressions net pennies.
While at times, ad-tech vendors take a huge cut of transactions, the study found that specifically supply-side platforms (SSPs) who work with publishers at times take a loss on transactions. Franaszek suggested that by lowering their take rate to a negative percentage on some bids, SSPs may be able to bring in more revenue for publishers at times, juicing their overall win rate.
“It certainly matches my experience,” the publishing executive said, adding he has seen instances where the winning bid for inventory is significantly lower than the bid cap, potentially suggesting that an ad-tech company might be intervening. Though, without studies like this, it is impossible to prove what’s behind this dynamic, there will naturally be idiosyncratic relationships between vendors and publishers that explain some variance in take rates.
Supply-side platforms want to prove that their technology brings publishers the most value and revenue. While some do add value, others prove added revenue through outcompeting other ad-tech vendors, which the publishers are already using, rather than delivering differentiated value, said Paul Bannister, chief strategy officer of publishing network Café Media.
“It doesn’t help the publisher in total,” Bannister said. “It’s just shuffling money from place to place.”
Premium publishers don’t always get premium rates
The study found several instances where made-for-advertising sites—those with articles like “49 photos of royals at the beach”—are charged a lower transaction fee from SSPs than marquee arts and literature publishers, reputable political news sites and business new sites. This effectively lets premium news publishers keep less of their ad revenue.
The reason, the study suggests, is that ad-tech platforms prioritize impression volume over quality.
But also, clickbait sites “were the first to really get smart and negotiate with [SSPs],” said Ali Manning, COO at Chalice, an ad-tech company that helps advertisers optimize their supply path. “They likely have a lot of negotiating power.”
On the flip side, premium publishers are, overall, better positioned to take advantage of the ad-tech supply chain, said Eric Wheeler, CEO of SSP 33Across.
“They have more clarity to how they’re getting paid and the path to their monetization than ever before,” said Wheeler, who also questioned some of the study’s conclusions. “It’s hard to believe that the premium publishers are being taken advantage of.”
Sources recognize the limits of the study. Cafe Media’s Bannister pointed out revenue sharing is more complicated than the research presents, though he agreed with most of the findings.
And any analysis of the vagaries of ad tech is incomplete without considering Google’s market dominance.
“The auctions that happen are pretty opaque so there’s potential for manipulation on both sides,” Manning, a former Google exec, said. “Google’s market influencer could be a big reason for the variation in take rates.”
Franaszek said the preliminary, exploratory nature of the study prohibited controlling for all confounding variables.
For the first publisher exec source, reports such as these are illuminating but frustrating due to a lack of resources, regulation and accountability.
“[Ad-tech vendors] are incentivized to take as much money [as they can],” the source said. “There’s very little ability to catch them doing so, and there seem to be very little consequences if they do.”
