In March, on a stage at The Glasshouse in New York, Jeff Green was asked about Amazon. He had an answer ready, because he always has an answer ready. Amazon's time in the open internet, he said, would be shorter than Google's.

Their demand-side platform was not really a demand-side platform. And the verdict, delivered with the calm of a man who has never once been wrong in public: "There isn't enough upside in their DSP as a standalone."

Five months later The Trade Desk told investors that next quarter's revenue would be roughly 12 percent smaller than last year's.

That is the whole story, and everything below is just the arithmetic.

Second quarter revenue came in at $715 million, up 3 percent, against a consensus of $752.6 million. Three percent is the slowest growth in the company's history as a public company, pandemic quarter included. GAAP net income fell 29 percent to $64 million. Adjusted EBITDA margin compressed from 39 percent to 34.

None of that is what took the stock down 24 percent after hours to $13.39, a price the company has not seen in seven years. What took it down was the guide: at least $650 million for the third quarter, against $739 million in the same quarter last year, and against a Street model of $804 million. Adjusted EBITDA guided to roughly $160 million, a margin near 24 percent against 43 percent a year earlier.

Green's framing was that the numbers "are not a reflection of our company." The market took about eleven minutes to disagree.

Two reasons, in this order. Macro first: consumer packaged goods and automotive, which together supply about a quarter of platform spend, are under pressure. Then, in one sentence against several minutes of Procter and Gamble and General Motors commentary, execution.

Analysts asked four separate times for the split between the two. They did not get a number. For a company built entirely on the proposition that it can measure what everyone else cannot, declining to quantify its own miss is a choice with some irony in it.

The mechanism Green described is more interesting than the excuse. Brands under pressure, he said, are not just cutting. They are moving to what he called low-cost, low-decisioning methods, and he named programmatic guaranteed and fixed-price deals specifically. Buyers taking that route, he said, hand their decisioning to sellers in exchange for cheaper transactions. He called the approach "often deliberately short-sighted."

Read that again, because it is a confession dressed as an insult. The CEO of the largest independent decisioning platform on earth just told the market that when money got tight, his customers decided decisioning was not worth paying for.

Pressed by New Street's Dan Salmon on Amazon's exclusive sports inventory and Google's buyer-direct program, Green ran the fee math on air. If The Trade Desk charges 8 percent and a competitor charges 4, he argued, you cannot conclude The Trade Desk has to be twice as good. Add the media back in and it is $1.08 against $1.04, and the question becomes whether the better impression was worth four cents.

It is an elegant argument. It is also built on numbers he made up on the spot, because The Trade Desk does not disclose its take rate and never has.

Trade reporting has put Amazon DSP at roughly 1 to 2 percent of media spend, with Amazon charging 1 percent on open web publisher inventory and nothing at all on programmatic guaranteed deals against its own media. The same reporting puts The Trade Desk in the 12 to 15 percent range. Run Green's own comparison with those figures instead of his and it is not $1.08 against $1.04. It is $1.12 against $1.01, and the performance gap the platform has to deliver to justify itself roughly triples.

Asked by KeyBanc's Justin Patterson about pricing, Green was direct: "It is not our goal to be the cheapest platform."

He is going to get his wish.

He also delivered the most quotable line of the call, on why decisioned buying beats the alternative: buy media at random, "you will get your ass kicked every single time." Nobody asked the obvious follow-up, which is that programmatic guaranteed is not random. It is the opposite of random. It is a named publisher, a fixed price and a known unit, which is what buying looked like before the industry invented a fee for pretending it was complicated.

Here is the thing Green's cheap-media framing cannot survive contact with.

On my own show two year ago, Mike Zaneis of TAG described what happens when you compare private marketplaces against the open exchange. He found "virtually no difference in invalid traffic, fraud rates, viewability rates." One variable moved. PMPs cost more. His question at the time was whether buyers were getting what they paid for.

Buyers spent the next two years answering it. When a CFO squeezes the budget and the media buyer goes looking for what to cut, the first thing to go is the line item nobody has ever been able to prove. That is not short-sightedness. That is a priced verdict on a value claim that never got demonstrated, and The Trade Desk conceded on this very call that the measurement product capable of demonstrating it is still in alpha, in the hands of very few customers, with no partners named and no ship date.

The company is asking the market to fund the proof of a proposition it has been charging for since 2016.

The most useful number in the release is the one nobody led with. Customer retention stayed above 95 percent, as it has for over a decade, while revenue grew 3 percent.

Those two facts together describe a specific condition, and it is not churn. Nobody is firing The Trade Desk. They are keeping the seat and spending the incremental dollar somewhere else. That is a much harder problem than losing accounts, because it does not show up in any metric the company reports until it shows up all at once in the guide.

Which it just did.

In January 2021, on a Variety panel at CES, Linda Yaccarino explained why NBCUniversal put its premium inventory into connected TV through The Trade Desk. She described the requirement as an "open marketplace, fully transparent," and said companies without that discipline were in the short-term commodity business.

In March 2026, on the Marketecture stage, Green said his company had at times been "way too open," to the point that it was not practical transparency. He compared The Trade Desk's invoice to a restaurant handing you a list of every ingredient you just ate, after which everyone argues about the price of croutons.

Between those two dates: Dentsu and WPP walked away from OpenPath in February 2026 over transparency and fees. Publicis told clients to stop transacting on the platform in March after an audit found the company had applied its DSP fee on top of other fees and enrolled clients in fee-based products without documented authorization.

Asked about the agency relationship on last week's call, Green said there were public disputes about what was essentially a negotiation, "but that's behind us." He added that all holding companies are in periods of transition.

At Marketecture in March, asked directly about Dentsu and WPP leaving, he said he would not comment on those two specifically. He still has not. Two of the largest buyers of media on earth walked out of his supply path and the company's public position remains that this is not a thing that happened.

The most honest sentence on the call came from Green describing why the guide looks the way it does. The company is building room into the forecast for its newly hired leaders to get up to speed and make a meaningful contribution.

That is a company telling investors, in writing, that next quarter's number assumes its own commercial organization will continue to underperform while it learns the business.

The organization in question: a chief financial officer who is the fourth in fourteen months and was presenting his first quarter. A chief commercial officer hired in July. A chief marketing officer, a chief business development officer, a chief operating officer and a VP of client strategy, all recent. Two new board members, one of whom arrived ahead of a Nasdaq independent-director cure deadline.

Green's closing line was that days like this one make the comeback story more impressive. The first measurable test of that arrives in early November, and the bar he set for it is $650 million.

How this was reported: revenue, margin, guidance and balance sheet figures are from The Trade Desk's Q2 2026 press release, Form 10-Q and Form 8-K filed August 6, 2026, and share prices from public market data through August 10. Quotations from the August 6 earnings call are drawn from a published account of the call and are being confirmed against the company's own recording before publication; readers should treat any wording not yet confirmed as approximate. Quotations from the March 2026 Marketecture Live III session and the January 2021 Variety CES panel are from transcripts of publicly published recordings and carry the same caveat. The Zaneis remarks are from ADOTAT's own February 2024 recording. Fee-range figures for Amazon DSP and The Trade Desk are trade-press estimates, not disclosed figures; The Trade Desk does not publish its take rate. The Trade Desk was contacted before publication and declined to reply. ADOTAT has no financial relationship with The Trade Desk, Amazon, Publicis, Dentsu or WPP, took no payment or advance review from any party named here, and is funded by reader subscriptions.