The Trade Desk's Bumpy Ride Down

Full analysis

The Trade Desk lost about 90% of its value in roughly seven quarters, cut 15% of its staff, and churned through most of its C-suite, including something like four CFOs in 18 months. Richard Kramer of Arete Research walked AdExchanger's James Hercher and Sarah Sluis through the wreckage. What the fall says about whether an independent DSP can hold the middle of the market at all, now that Amazon, Google, and the agencies themselves are all pulling in the other direction, is the question worth sitting with here.

How hard is this to undo? For The Trade Desk, hard. Take rate erosion, holdco defections, and a rebuilt leadership team are not fixed in a quarter. For everyone reading this, the useful decision is easy to undo: how much of your roadmap and your budget you route through a single independent buying platform.

What's actually being decided: whether the "neutral independent DSP" position is a durable business or a phase we're now exiting.

The Market Analyst. Kramer's three-part indictment is the meat, and two of the three are structural, not cyclical. Analysts priced recurring campaign spend as if it were subscription software revenue that renews itself. It doesn't. Campaigns stop. Second, the take rate sat at ~20% for years while The Trade Desk bolted on Open Path, UID 2.0, and data marketplace fees that should each have lifted it. Flat take rate plus new fee products means take rate was falling somewhere and getting papered over elsewhere. The Trade Desk killed gross-spend disclosure in 2021, so nobody outside could check. In plain terms: the company stopped showing the denominator right before the ratio got interesting.

The Skeptic. Kramer is a long-time bear, and he has an incentive to make the antitrust ruling sound worse for The Trade Desk than it is. His claim that Google's network business is "less than 1% of profits" is his own framing, and it conveniently supports his argument that the behavioral remedy is toothless. Fine. But the load he's carrying holds up on the parts that are checkable: holdcos really did defect, Amazon really did win the Omnicom account and lock up Disney, Netflix, and Roku supply, and Green really did keep denying Amazon was a competitor. When your CEO tells investors the thing eating your growth isn't real, and it is, that costs you the benefit of the doubt at exactly the wrong moment.

The Operator. Here's what breaks Tuesday morning. Four CFOs in 18 months means nobody in that building can answer a hard forecasting question from memory. Kramer's point on the layoffs is the one operators feel in their gut: the 15% who left were not doing nothing, and their work now lands on a senior team that's still learning where the bathrooms are. Recovery gets slower before faster. And the holdco moves aren't abstract. Publicis telling clients The Trade Desk failed an audit is a knife in a renewal conversation. Dentsu and WPP leaving Open Path means the direct-to-publisher pipe The Trade Desk built to defend its position is leaking on the demand side.

The Customer / End User. The customer here is the agency, and the agencies have already voted. They're building their own identity spines, which means they no longer need to rent LiveRamp or ID5 through The Trade Desk's marketplace. That removes a fee layer The Trade Desk was counting on and removes a reason to route through it at all. When Omnicom won Amazon and moved spend to Amazon DSP, that wasn't a betrayal, it was math: Amazon has the conversion data lake and the CTV supply deals. The "white knight of the open internet" pitch was a great story on the way up. On the way down it just tells four holdcos exactly whose growth they'd prefer to own instead.

Where the council splits. Kramer thinks Google gets more dangerous now that the antitrust case is over, because DV360's team no longer has to assume every Slack message is a discovery exhibit. The Skeptic's read is that a company sitting on a monopoly finding, watching the remedy phase, does not suddenly get reckless with its buying-side aggression. Both can't be right. The second real split: is The Trade Desk's problem the CTV margin mix (structural, permanent) or the execution stumble around Kokai and guidance (fixable with a steadier hand)? If it's mix, no CEO saves it. If it's execution, the new team eventually does.

What it hinges on. Two beliefs. One: whether a buying platform can charge ~20% for sitting between advertisers and inventory when the two biggest inventory-plus-data owners, Amazon and Google, can undercut that fee because ads are a rounding error against their real businesses. Two: whether agencies, now that they own their own identity plumbing, ever again want a neutral middleman badly enough to pay for one. The council leans hard toward "structural." The M&A regret Hersher raised, buying Criteo for retail media or Roku for supply and first-party data when the market caps allowed it, is really an admission that The Trade Desk needed to own data or inventory and chose to own neither.

What to de-risk if you're an operator: don't hard-wire your stack to any single independent DSP's identity or supply path. The Open Path defections show how fast that pipe can narrow. Keep a live Amazon DSP and DV360 integration whether or not you love them, because your holdco partners already do.

Prediction: The Trade Desk's full-year 2026 revenue growth, reported on its Q4 2026 earnings call in February 2027, will come in below 15%, confirming this is a repricing of the independent-DSP model and not a one-quarter stumble.

Confidence: Medium. The demand-side leaks are structural, but a soft comparison base could flatter the number.

Why: The damage Kramer describes is on the demand side and it compounds: Publicis warning clients off, Omnicom moving to Amazon DSP, Dentsu and WPP leaving Open Path, and agencies building their own identity infrastructure so they stop paying The Trade Desk to resell it. Those are contract and workflow decisions that don't reverse inside a year, and they land on a leadership team that's still rebuilding after near-total C-suite turnover and a 15% cut. The bull case requires the CTV growth story to outrun holdco defection and a newly unleashed DV360, and there's nothing in this teardown that says it will. The opposite outcome, a snap back above 20% growth, would require the same four holdcos that just walked to walk back, which nobody in this conversation thinks is happening.

Revisit by 2027-03-01: We're right if The Trade Desk reports full-year 2026 revenue growth under 15% on its Q4 2026 call. We're wrong if it reports 20% or higher, or explicitly reverses one or more of the holdco defections (Omnicom spend returning from Amazon DSP, or WPP/Dentsu rejoining Open Path).

One more thing worth carrying out of this episode past the schadenfreude: the MCP interface layer Mark McEachran described cuts both ways for independents. If buyers can issue natural-language budget-shifting commands across Google, Meta, Amazon, and every DSP through one connector, the switching cost that protected a platform like The Trade Desk gets thinner, not thicker. Easy to move spend is good for whoever owns the data and the inventory. That's not the independents.