The Amazon-FTC case and the myth of ad auction transparency

Full analysis

Your draft

The FTC sued Amazon over how its sponsored-search auction actually charges advertisers. Amazon quietly switched from a second-price auction, where the winner pays a hair above the second-highest bid, to a hybrid that adds a hidden floor. When Amazon decided an impression was worth more, it charged the winner above the runner-up's bid and told nobody. Tim Peterson and Seb Joseph, on the Digiday podcast, land on a bleaker point than the lawsuit: nobody buying ads seems to care, because the results still show up.

That is the decision on the table for operators. Not "will Amazon lose." The real question: if auction transparency is dead and buyers have stopped asking for it, what is left to sell, and to whom?

This is easy to undo in the sense that no operator has to bet the company on it this quarter. It is hard to undo in the sense that once platforms own targeting and auction mechanics, and creative next, there is no getting those levers back. The deadline is set by the platforms' roadmaps. The courts are not driving this.

The Market Analyst. Two regulatory outcomes landed the same week and both point the same way. The Google remedy did not force divestiture of the ad server or the exchange, and the DSP was never even contested. Seb Joseph's read on Amazon: "I haven't seen enough examples of any of these tech platforms losing." Believe him. For anyone long independent ad-tech on a thesis that antitrust rebalances the market, that thesis is done. Regulators are not breaking the walled gardens. Peterson notes the DOJ actually did seek the exchange divestiture in its proposal, so the story is still moving, but the direction of travel is minimal structural change. Publicis gets a mention as the best-run holdco that still cannot move its stock. That is the market telling you execution does not beat structural compression. For a non-specialist: the referees just declined to break up the biggest players, and investors betting they would need a new bet.

The Skeptic. Everyone agrees transparency is dead, so let me push on it. The 70% number is doing a lot of the work here. One programmatic buyer told Peterson that 70% of their log-level data, the impression-by-impression records DSPs hand back, doesn't even say whether the auction was first- or second-price. That is not a claim about Amazon. That is a claim about the whole open programmatic pipe. If it holds, then auction-strategy optimization was always mostly theater, and the buyers who sold it as a service were selling a lever they could not actually pull. But one buyer is not a market. Nobody has audited that 70% across DSPs. And "buyers have moved on" is convenient for the platforms and convenient for the agencies who never optimized in the first place. Resignation is not the same as consensus that nothing was lost.

The Operator. Try to act on this Tuesday morning and you hit a wall fast. You want to prove Amazon overcharged you? You need the auction data Amazon controls. You want to shift budget away in protest? Performance is still hitting ROAS targets, so your client says no. Google's August 17 change tightened Smart Bidding to hit the exact return-on-ad-spend target an advertiser sets, which kills the occasional happy accident where a good buyer underpaid for a converting impression. Both the downside and the upside of skill are being engineered out. What breaks at 90 days: your pitch. When targeting and auction are automated, the deck that sold optimization has nothing in it. The independent agency CEO who told Joseph at Cannes he is turning his shop into a boutique consultancy is not being dramatic. He is reading the P&L.

The Customer / End User. Here the customer is the brand and the agency buying the media. They are behaving rationally and it looks like surrender. Every agency exec Peterson talked to shrugged, because the number they are graded on is still green. That is the whole game. Advertisers never came to Amazon, Meta, or Google for pure auctions. They came for results. So the platforms can change the mechanics under the hood and face no revolt, as long as ROAS holds. The uncomfortable part Peterson names: agencies feeding first-party data, Google Analytics, attribution models, into these platforms are training the systems that replace them. "At what point are you just training your replacement?" The customer is helping build the thing that eats the customer.

The CFO. Follow the money and creative is the only line item with margin left. Peterson calls it the last lever, the one variable platforms cannot yet fully account for. Meta is hoovering up creative variants under the banner of optimization help, which Joseph frames as training data collection for a generative system that will eventually make the creative itself. Zuckerberg has said publicly, per a Wall Street Journal interview Joseph cites with a hedge, that AI creative is a priority. So the vendors with real runway are the ones sitting in the creative layer: dynamic creative optimization, generative ad tools, creative analytics. But price that runway honestly. If the platform end state is budget-in, KPI-in, results-out, then creative automation is the platforms' next acquisition target, not a durable moat for anyone else. You are renting the last room in a house the landlord plans to knock down.

Where they part ways. The Skeptic and the Customer disagree on whether the shrug is informed. The Customer says buyers made a rational trade: mechanics for results. The Skeptic says buyers never had the data to know what they traded away, so calling it a choice flatters everyone. Second split: the CFO sees creative as the value that survives; the Operator and the Market Analyst see it as the next thing absorbed, on a clock set by Meta's model quality. If creative is a two-year moat, invest. If it is a two-year lease, harvest and move upstream now.

What this actually hinges on. One belief: will regulators change market structure? The council says no, and the same-week Google remedy is the evidence. Second belief: is creative a durable lever or a temporary one? Nobody on the podcast thinks it is durable past the point Meta's generative creative gets good enough. That is the whole call. If you are a measurement or identity vendor, the opening is not transparency as a moral crusade, because buyers stopped paying for it. The opening is verification that the automated systems are actually delivering what they claim, which is a different and more sellable product. Before committing real budget to a "transparency" pitch, test whether even one large buyer will pay for it. The episode suggests they won't.

Prediction: No major ad-tech or measurement vendor will build a commercially successful "auction transparency" product off the Amazon FTC case, and the case itself will end in settlement or minimal terms with no forced change to Amazon's auction disclosures, by the FTC's expected 2027 trial-scheduling milestone.

Confidence: Medium. Buyer indifference is documented, but litigation timing can surprise.

Why: The signal in this episode is that every agency exec Tim Peterson interviewed shrugged at the Amazon allegations because their ROAS numbers are still green, and one programmatic buyer says 70% of DSP log data doesn't even reveal auction type. A product only sells if buyers will pay, and these buyers have stopped asking for auction transparency, so there is no demand to build against. The mechanism that would prove the opposite, a wave of advertisers pulling budget or demanding auction audits, requires them to act against the results that currently satisfy them, and nothing in the episode suggests that is coming. The Google remedy landing the same week with no divestiture of the ad server or exchange is the template Joseph and Peterson both expect Amazon to follow.

Revisit by 2027-06-30: We're right if the Amazon FTC matter is settled or narrowed with no mandated change to how Amazon discloses its sponsored-search auction, and no independent vendor has launched a transparency product with named paying buyers. We're wrong if Amazon is forced by ruling or settlement to disclose its auction mechanics to advertisers, or a DSP/SSP/measurement vendor publicly signs multiple large buyers to a paid auction-transparency product.