Are DSP fees holding back indie agencies?

Full analysis

The Trade Desk and Amazon are gatekeeping their demand-side platforms (DSPs, the software advertisers use to buy ads programmatically) against small independent agencies. Indies get routed through resellers, stack fees, and wait on emails that holdco agencies never have to send. Digiday's Kimeko McCoy and Tim Peterson diagnosed it well. The real question for operators is whether this is a durable moat or a resource squeeze that just created an opening for everyone else.

This is a business-model choice, and those are easy to undo. The Trade Desk can lower a spend threshold with a memo. Which means the moat here is thinner than it looks. What's actually being decided is who serves the long tail of ad buyers, and whether that tier is worth serving at all. Nothing sets a hard deadline, but the AI-agent angle McCoy and Peterson raised is the clock: once buying happens through Claude or ChatGPT instead of a human account manager, the whole gatekeeping logic changes.

The Market Analyst. The $100,000-to-$300,000 monthly spend floor tells you The Trade Desk is rationing a scarce resource, and that resource is people. Peterson tied the gatekeeping directly to The Trade Desk's recent layoffs and executive turnover. Fewer account managers means you triage toward Publicis, not a five-person shop. For an operator reading this: The Trade Desk is protecting margin by refusing low-margin customers. That is a company managing cost, and it reads as a company under pressure, not one expanding. McCoy's line that the number of clients demanding The Trade Desk by name is "dwindling" is the part that should worry Trade Desk holders. The lock-in is loosening before a replacement growth lever exists.

The Skeptic. Before everyone cheers the AI-agents-save-the-indies story, ask what has to be true. The thesis is that MCP servers (model context protocol, the plumbing that lets an AI assistant talk to another system) let agencies route buys through an AI and skip the human account manager. Meta already opened its ad API to Claude and ChatGPT, so it's live somewhere. But The Trade Desk gatekeeps direct seats to protect fee economics, not just to save on support headcount. An agent doesn't lower the 17% platform fee. It removes the reseller's 12%, which helps the indie and hurts the reseller, and does nothing to The Trade Desk's take. The agents level service, not price.

The Operator. Picture the indie agency lead Tuesday morning. Client insists on The Trade Desk by name because it performs. You can't get a direct seat, so you're paying 29% all-in through a reseller against 17% direct. McCoy reported at least one agency filling a quarterly spend gap with its own money to hit a threshold. That is a shop subsidizing a vendor's minimum out of pocket, and it ends in losses or accounting gymnastics. What breaks first is the client relationship when performance drops because 12 points of fee got eaten before a single impression served. The escape route everyone mentions, StackAdapt or Yahoo or DV360, is real but inferior. Google's DV360 is expensive and needs an inside connection to even access.

The Customer / End User. The client here is the advertiser, and the advertiser doesn't care about seat politics. They want The Trade Desk because it works, and they are starting to stop caring. That's the shift. When a client stops naming The Trade Desk, the whole access fight becomes moot, because the indie just moves the budget to a walled garden. Meta, Google Search, TikTok, and now the AI chat surfaces don't make you qualify for a seat. You open a self-serve account and spend. The open web requires a DSP relationship. Meta requires a credit card. For a cost-strapped indie serving a performance client, that math is not hard.

The CFO. Both Amazon and The Trade Desk are, in Peterson's words, leaving money on the table by not building a low-cost self-serve tier for small buyers. True, but only if that tier is profitable. The reason it doesn't exist is that the support cost of a small account swamps the revenue. That's the actual constraint. An AI-agent self-serve layer changes the equation because it removes the human cost, which is exactly why it's attractive to a DSP watching headcount. The payback is real: serve the long tail at near-zero marginal support cost, recapture the fees currently going to resellers like Mogul. The question is whether The Trade Desk builds it before an SSP moving up the stack does.

Where the council splits

Two real disagreements. First, is AI-agent access a leveler or a mirage? The Operator and CFO see a genuine path to cheap self-serve. The Skeptic says it fixes service but leaves the platform fee intact, so the headline 17% that squeezes indies barely moves. Both can be right: the reseller layer dies, the DSP take survives.

Second, does the indie tier even matter? Peterson argued the open web is eroding under bot traffic he put above 50%, a number the fact-check flagged as high and unsourced. If he's directionally right, the indie fight is over inventory that's shrinking anyway, and the budget migrates to walled gardens on its own. If he's wrong, the open web is a real demand source that The Trade Desk is actively neglecting, and someone will take it.

What it hinges on

It comes down to one belief: is small-buyer demand worth serving? The Trade Desk is betting no, at least not at current support costs. That bet is only safe if two things hold. Clients keep insisting on The Trade Desk by name, and no cheaper serviceable alternative appears. Both are weakening at once. McCoy's sources say the name-insistence is fading, and the SSPs, Magnite, PubMatic, and Index Exchange, are already pitching the buy side directly. The council leans toward this being an opening for challengers, not a durable moat. The thing to verify before acting on it: whether an AI-agent buying layer actually clears at low enough support cost to make the long tail profitable, or whether it just shifts fees around without changing the unit economics.

Prediction: At least one supply-side platform among Magnite, PubMatic, and Index Exchange will publicly launch a self-serve or direct buy-side product aimed at small and independent agencies by the close of Q3 2026 earnings season in November 2026.

Confidence: Medium. The convergence is already underway, and The Trade Desk's visible neglect of the indie tier is the trigger.

Why: Peterson described SSPs already pitching buy-side clients directly while The Trade Desk rations seats behind a $100,000-plus monthly floor and cut the account-management headcount that would serve smaller buyers. An SSP that already runs the publisher-side pipes can offer indies direct access to inventory without the reseller's 12% stack, and it has every incentive to grab a demand segment the market leader is openly turning away. The opposite outcome, SSPs staying purely supply-side, is the less likely one because the "unified ad platform" land-grab is the stated strategy for all three and the indie tier is the cheapest demand to win right now. The one thing that would break this: if the AI-agent buying layer arrives first and commoditizes the pitch before an SSP can brand a product around it.

Revisit by 2026-11-30: We're right if Magnite, PubMatic, or Index Exchange announces a self-serve or direct buy-side offering targeting independent or small agencies by the end of November 2026 earnings season. We're wrong if none of the three launches or announces such a product in that window.