DSP Wars Episode III: Revenge of the Holdcos

The programmatic advertising industry had a very public moment the last few weeks. Publicis — one of the world’s largest advertising holding companies — told its clients to stop using The Trade Desk, citing a failed audit over fee transparency and unauthorised billing. The Trade Desk denied the characterisation and its CEO Jeff Green published a Linkedin post calling out agencies who wave the flag of transparency publicly but run from it in practice.

It was messy. It was pointed. And it was almost certainly not really about the audit. This is a story about the structure of the programmatic market — who owns the infrastructure, who profits from it, and whether that structure is capable of producing good outcomes for advertisers regardless of the intentions of the people inside it.

To understand what is actually happening, you need to see that what looks like a single product category — the DSP — is actually three fundamentally different business models wearing the same software clothing.

The first is the pure-play DSP. The Trade Desk is the primary example. Revenue comes from a percentage of media spend — typically 15-20% in platform fees. The model is structurally aligned with advertiser outcomes: if campaigns perform and you buy more, they make more money. There is no proprietary inventory to push, no principal position to protect, no inventory owner sitting on the other side of your auction. In 2024, The Trade Desk processed approximately US$12 billion in gross spend while maintaining customer retention above 95% for eleven consecutive years. You do not hold that retention figure in a market as competitive as programmatic buying by systematically working against client growth incentives.

The second is the media-owner DSP. Google’s DV360 and Amazon DSP are both built by businesses that own enormous quantities of the inventory you are buying through them. Google operates the largest digital advertising inventory in the world. Amazon owns Prime Video, Twitch, and a commerce data infrastructure unmatched outside the walled gardens.

Both offer third-party inventory too. But the economic incentive of an infrastructure owner who also controls supply is not the same as a platform whose only job is to find you the best media in the market. This is not a criticism of the quality of either platform’s inventory — it is a description of the structure they operate within and can’t economically ignore. They can price at rates very different to the major DSPs because they are also monetising other parts of the chain.

The author Henry Innis

The third is the holdco DSP. GroupM’s Nexus, Publicis Media Exchange, and Omnicom’s programmatic practice sit inside agency structures that not only buy media on your behalf — they increasingly hold principal media positions. A principal media position means the agency bought inventory for itself, at scale, and is now reselling it to you. Think of it as a stockbroker buying shares for their own book and then recommending you buy them. The conflict is structural, not personal. It does not require anyone to act in bad faith to produce outcomes that are misaligned with your interests. It just requires the incentives to be what they are.

Every DSP presents the same dashboard, the same targeting parameters, the same report. But the economics underneath are fundamentally different.

The question is why the holdco market is moving in this direction, and to answer that you need to think about it like a banker rather than a marketer.

WPP currently trades at an EV/EBITDA multiple of around 5.7x. Omnicom at approximately 6.9x. These are mature services business multiples — the kind of number you get when a market concludes your revenue is defensible but not growthy, your margins are structurally constrained, and your core value proposition is relationships rather than technology.

At its peak in December 2024, The Trade Desk carried an enterprise value of $56 billion on revenues under $3 billion — a revenue multiple of nearly 20x. Even after its recent decline, the valuation still reflects the significant premium the market gives to technology infrastructure over services businesses.

The inexorable push to platform

That gap is not lost on the holdcos. Building a DSP, or building a principal media trading desk that behaves like one, is not just a services decision. It is a financial engineering decision. The fastest path to re-rating a holdco’s equity is to make it look less like an agency and more like a platform. This does not require any conspiracy or malice — it is the rational response to a valuation structure that rewards platform economics over services economics. Any management team in the same position would do the same thing. The problem is not that holdcos are acting irrationally. The problem is that the rational action for their equity story and the optimal action for your media plan are not the same thing.

If this structural conflict sounds familiar, it should. Financial regulators spent decades constructing rules that separated exchange operators from proprietary traders, precisely because the same dynamic had played out in capital markets. The principle they landed on was straightforward: if you control the infrastructure through which trades are executed, you cannot simultaneously use that infrastructure to trade your own book. The conflict is too fundamental to manage with disclosure alone.

Barclays paid $70 million and Credit Suisse paid $84 million to settle allegations that they misled investors about how they managed their dark pools — private trading venues where they were simultaneously operating the infrastructure and participating as principals. Michael Lewis’s Flash Boys documented in detail how this structural conflict systematically disadvantaged ordinary investors. The book became a bestseller not because the mechanics were complicated, but because the principle was instinctively obvious once explained: when someone controls the pipe, they know what flows through it before you do.

The important part of this history is not the fines. It is that regulators concluded the conflict was structural and required structural separation — not just better intentions from management or more frequent audits. The programmatic advertising ecosystem has the same structural problem. A 2025 report by the World Federation of Advertisers found that nearly half of advertisers can only audit their programmatic spend through the signing agency, with more than half lacking clear penalties for non-compliance with principal media workflows. The infrastructure has evolved faster than the governance around it. That is not an accusation. It is a diagnosis — and it is one that matters regardless of which holding company you work with.

This is why the current pressure on The Trade Desk — multiple holdcos reducing reliance on independent platforms as they accelerate their own proprietary trading infrastructure — should concern every advertiser, not just those directly affected.

An independent, scaled, conflict-free DSP is not just good for the advertisers who use it. It functions as a price reference for the entire market. When an independent platform wins a campaign against a holdco trading desk, it creates a data point about what programmatic buying actually costs when nobody is extracting a principal margin. A market in which holdco trading desks face no credible independent benchmark is a market in which those trading desks set their own prices.

Independence as a virtue

Inside the Trade Desk officer in Chicago

What does that independence look like in practice? Two examples are worth understanding — not as product endorsements but as illustrations of what structural independence makes possible.

First, supply path transparency. The Trade Desk built OpenPath to give advertisers a more direct route to premium publisher inventory, reducing the intermediary hops — and intermediary fees — between the buyer and the publisher. When you buy a premium news impression through a vertically integrated trading desk, there may be multiple layers of margin extraction between your dollar and the publisher’s page. OpenPath collapses that chain and publishes the path so you can audit it. A platform that profits from the opacity of that chain has no commercial incentive to build the equivalent. An independent platform does.

Second, cross-channel frequency management. A vertically integrated trading desk typically manages frequency — how many times a single person sees your ad — within its own inventory positions. An independent platform operating across open exchange inventory with no proprietary position to protect can manage frequency holistically across publishers, formats, and devices. For a brand running across streaming TV, display, and audio simultaneously, this means a consumer is not hit seven times on one channel while being ignored on another. That is better reach efficiency and lower waste — a direct benefit that a platform with inventory commitments has structural reasons not to optimise for.

Holdco platforms are not without genuine strengths. Scale in media buying translates to inventory pricing leverage that no independent platform can match in direct negotiations. Operational integration with planning, creative, and analytics teams has real value. For many advertisers, the convenience and buying power of a vertically integrated stack will outweigh the structural concerns described here — and that can be a legitimate, informed choice. The issue is not that holdco platforms should not exist. It is that they should not exist without an independent alternative functioning as a market benchmark.

None of this is to say The Trade Desk is beyond criticism. At its scale, pricing complexity is a genuine operational reality. Its expansion into strategy and direct client engagement has made it a competitor to the agencies that built it into a category leader, and that tension is real. But the structural argument for why an independent DSP of significant scale needs to exist is entirely separate from whether The Trade Desk is executing perfectly right now. The category is more important than the current leader within it.

What CMOs should actually do

Two things follow from everything above, and both belong in your next quarterly media review.

First, demand platform independence as a structural principle. Your primary buying platforms should not be owned by the people selling you inventory or holding principal positions in the media you are buying. If you are using the same entity as your media agency, your trading desk, and your DSP, you are asking one vendor to perform three distinct commercial functions with three distinct sets of incentives. The fact that they are bundled together does not make the conflicts disappear — it makes them harder to see. This is not anti-agency. Agencies do important, skilled work. It is pro-accountability — and it is the same principle financial regulators arrived at after decades of learning the hard way.

Second, invest in measurement that sits above the DSP stack entirely. The only way to evaluate what any DSP is actually delivering — net of incentive structure, net of pricing complexity, net of inventory conflict — is to measure outcomes from a layer that has no financial relationship with any of the platforms being evaluated.

This is what marketing mix modelling does. A well-constructed MMM — built on media data, economic factors, pricing, business outcome data, and a correctly specified model — does not know or care whether your spend went through The Trade Desk, DV360, or a holdco principal media desk. It measures the revenue contribution of each channel, each platform, and each campaign in the context of everything else happening in the market. It is the only measurement layer that is structurally independent of the infrastructure being measured.

I’m obviously not a neutral observer here. But the more important point is this: the independent MMM sector is large, growing, and full of serious operators. Analytic Partners has been doing this for 25 years. Kantar operates across markets globally. There are others — Recast, Nielsen, established consultancies — all building on the same structural premise: that measurement sitting outside the media supply chain will always be more trustworthy than measurement sitting inside it. The specific vendor matters less than the principle of having an independent system monitoring the system if there are financial conflicts within the system.

The structural conflict in media and programmatic is not going away. The holdcos are not acting irrationally. The market is rewarding them for moving in this direction. The only effective response is measurement that is genuinely independent — and the good news is that independent measurement has never been more available, more accessible, or more capable than it is right now.

Episode III ends with the Empire consolidated. The Death Star is almost operational. The galaxy does not realise what has happened until it is too late. The advertisers who avoid that ending will be the ones who understand that identical-looking surfaces can conceal fundamentally different incentive structures — and who build measurement infrastructure independent enough to see through the surface to the economics beneath.