Amazon's $20 Billion Auction Surcharge Is Every DSP Buyer's Problem Now
The FTC and 22 states sued Amazon on August 31, 2026, alleging Sponsored Products advertisers paid their full bid about 80% of the time in 2024 despite Amazon calling the auction second-price. The complaint covers over 1 million advertisers and an estimated $20 billion in extra charges since 2019. For DSP buyers, the number to audit is paid CPC divided by max bid.
I want to be upfront about my take here. The lawsuit itself is an Amazon story, and Amazon will litigate it for years. The useful part for the rest of us is smaller and more boring: every buyer now has a benchmark for spotting a first-price auction hiding behind a second-price label, and a good reason to ask every platform they buy through to describe its pricing rules in a document someone can be held to.
What the FTC says the "invented auction participant" did
The mechanism in the FTC complaint is simple once you strip the legal language. Amazon told advertisers the winner pays "one cent more than the next highest bidder." Starting around 2019, according to the FTC, Amazon added what it internally called a "soft reserve price," a floor that moved in real time and sat close to the winning bid. Internal documents quoted in the release call it an "invented auction participant" and a "clever non-transparent way to charge first price."
The trend line is the part I keep coming back to. Per the FTC, Sponsored Products advertisers paid their own bid 30% to 40% of the time in 2021, 70% in 2022, and about 80% by 2024. If that holds up in court, the auction did not flip overnight. It drifted, a few points a quarter, in a direction nobody outside Amazon could see from their reports.
The case was filed in the Western District of Washington and covers Sponsored Products, Sponsored Brands, and Display Ads. It does not, as far as I can tell from the filing summaries, cover Amazon DSP inventory. Hold that thought, because it matters for the second half of this piece.
Amazon's defense is better than the headlines suggest (and still dodges the question)
Amazon's response makes three points worth taking seriously. First, "in no scenario does an advertiser pay more than their bid," which is true and also the whole problem, since bidding your max in what you think is a second-price auction is textbook rational behavior. Second, roughly 92% of selected Sponsored Products ads were not the highest bid, because relevance gets weighted, and average winning bids fell 50% from 2019 to 2025. Third, average CPC stayed flat after inflation from 2019 through 2024 while conversion rates grew 24%.
On paper, that sounds like a platform that got more efficient for buyers. And in some ways it probably did. But none of those numbers answer the actual allegation, which is about the gap between what you bid and what you paid on the clicks you won. Flat CPCs and a 50% drop in winning bids are compatible with a world where the reserve ate most of the difference. Amazon's own line that advertisers "adjust bids based on real-world performance, not descriptions of auction mechanics" is, honestly, a strange thing to say in your own defense. It amounts to: you never knew the rules, so you couldn't have relied on them.
Emarketer's Zak Stambor told Marketing Dive the allegations "raise uncomfortable questions about how transparent Amazon is with advertisers," and he pointed out that leaving is not a realistic option when one platform controls roughly three-quarters of US retail media spend and cleared close to $70 billion in ad revenue in 2025. That is the bargaining imbalance the whole case sits on.
The ratio to pull from your reports this week
You cannot see Amazon's reserve price. You can see two things: what you bid and what you paid per click. Divide the second by the first, keyword by keyword, over the last 30 days. Sponsored Products search term and targeting reports give you average CPC, and your campaign settings give you the bid, so this is a spreadsheet exercise, maybe 20 minutes for a mid-sized account.
Then look at the distribution, not the average. In a functioning second-price auction with a reasonable number of bidders, a good chunk of your clicks should clear well below your bid. The FTC's figure gives you a rough tripwire: if most of your clicked keywords are sitting at 0.9 or above on that ratio, you are operating in something that behaves like a first-price auction, whatever the help docs call it.
I would treat that as a diagnostic, not proof of anything. Thin auctions on long-tail terms will naturally clear near your bid because there is nobody else there. The interesting signal is on your head terms, where competition should be pulling the clearing price down and, per the complaint, apparently was not.
The bidding consequence is the part most teams will skip. In a true second-price auction, bidding your real value per click is the correct strategy. In a first-price auction, that same strategy overpays by definition, and you are supposed to shade your bids below true value. If your account has been "bidding to value" since 2019 on the assumption the second-price description was accurate, the honest test is to shade your top 20 keywords by 10% to 15% for two weeks and watch two things: impression share and paid CPC. If CPC drops nearly point for point with the bid and impression share barely moves, the auction was clearing at your bid, and you have been paying for the label.
The AdExchanger piece on the case quoted Sarah Caputo of Fraction Method on exactly this: "An entire generation of bidding strategy was built on that one assumption, and most advertisers never had a way to check it themselves." Now you have a partial way. Use it.
Amazon describes the same auction four different ways, and that is the DSP story
This is where the story stops being about Sponsored Products. In a follow-up column, Caputo, who spent 22 years on the buy side, walked through how Amazon's own materials describe auction pricing across its ad business. The public PMP guide says a private auction is one where "the highest bidder wins the impression but pays a penny more than the second highest bid" and that "private auction deals are based on a floor price." DSP help documentation, per Caputo, describes Prime Video's Private Auction as a "variable CPM based on optimization." Third-party publisher inventory runs on a "unified, first-price auction." Elsewhere the same product is called "dynamic."
Four descriptions. One product family. No visible clearing formula for the one that matters most to CTV buyers.
I don't think that is evidence of anything improper on the DSP side. It may just be documentation drift across teams that never talked to each other, which is how most large companies work. But it means a buyer moving seven figures through Amazon DSP into Prime Video literally cannot point to a sentence that says what determines their price. And after August 31, "we assumed second-price because the guide said so" is no longer a defensible position to bring to a CFO.
To be fair to Amazon, opacity about auction rules has a long history in ad tech. Google moved its exchange to first-price in 2019 after years of confusion about what "second-price" meant when multiple auctions stacked on top of each other, and a former media buyer quoted by AdExchanger drew the comparison directly to the Google Search antitrust case. Anthony Vargas's framing was that the Amazon case is ad tech's history of opacity repeating itself. I think that is roughly right, with one difference: this time the buyers have a regulator's numbers to cite in the RFP.
Two questions to send in writing (and what silence tells you)
Caputo's recommendation is the one I would adopt as standard practice for every DSP relationship, not just Amazon. Send two questions, by email, to your account rep and copy whoever signs your contract:
- What auction methodology applies to each inventory type I buy through you? Open exchange, PMP, private auction, programmatic guaranteed, and owned-and-operated inventory each get their own answer. "Second-price" alone does not count. Ask what determines the clearing price when there is only one bidder, and whether any platform-set floor can move within an auction.
- Can I access log-level data that lets me verify that methodology? Bid, clearing price, and floor per impression, or at least per deal. If the answer is no, ask what aggregate report comes closest.
Give them five business days. From what I've seen of how these conversations go, you will get one of three responses. A written answer with specifics, which is great and now sits in your files. A phone call instead of an email, which is a soft signal that nobody wants the answer in writing. Or nothing, which is the answer. If you get the second or third, model your bids as first-price on that inventory going forward and shade accordingly.
Anyway, this also connects to a broader fee conversation. We wrote last week about how The Trade Desk's slide out of the S&P 500 made DSP fees negotiable again. Auction disclosure belongs in that same negotiation. A platform that cannot document how it prices your impressions has less standing to charge a premium for access to them.
Retail media's second-price honeymoon is ending
The Digiday explainer on the lawsuit ends with the question I expect to hear in a lot of Q4 planning meetings: does your second-price auction actually work the way you say it does? Walmart Connect, Instacart, Kroger Precision Marketing, and a dozen smaller networks all borrowed Amazon's auction language when they built their self-serve products. Most of them have never been asked to prove it.
My prediction, with the usual caveats: within 12 months at least two of the top five US retail media networks publish a written auction methodology document, mainly because agency RFPs start requiring one. The networks that do it early will get a small and probably temporary trust bump. The ones that don't will spend 2027 answering the question one client at a time.
And the sellers who were already withholding spend over rising costs, which CNBC noted had become an organized boycott among some top merchants, now have a document to point at. I don't know how the case resolves. I do know the paid-CPC-to-bid ratio is a column you can add to your report tomorrow morning, and once you've looked at it, you probably won't stop.
Notice Me Senpai Editorial
