Start with a media buyer holding a dollar. Before that dollar becomes an ad on a television, website or retailer app, it can pass through platforms that select audiences, compare inventory, run auctions, apply data and measure a sale. The screens may promise the same outcome: spend efficiently. The contracts tell a more interesting story.
The Trade Desk and Criteo are often placed in the same wide adtech bucket. Both give marketers software for buying digital advertising. Both talk about the open internet, data and measurable outcomes. Yet their disclosed business models are shaped differently. The Trade Desk says its clients are agencies, advertisers and service providers for them. It generally charges a platform fee based on a percentage of client spend, plus revenue from data and other services. Criteo sells demand-side tools to brands and agencies, while also selling supply-side monetization technology to retailers and media owners.
That difference is easy to miss because a product tour foregrounds controls, audiences and charts. It matters because revenue design is product design in slow motion. A company paid by a buyer has a direct reason to retain and grow that buyer's spend. A company connecting buyer and seller can create more liquidity, better data and simpler measurement, while carrying obligations to parties whose goals sometimes diverge.
The clean line The Trade Desk wants to draw
The Trade Desk's public argument is unusually plain: it does not own or operate media, so it can focus on buying inventory for clients. In its 2025 annual report, the company says it generates revenue from a platform fee generally based on a percentage of total client spend, alongside value-added services and data. It reported $13.39 billion of gross spend and $2.90 billion of revenue for the year.
Do not divide those first two figures and call the result a posted fee. The filing warns against that shortcut. Gross spend includes inventory, data, services and the platform fee. Revenue shifts with service use, client mix, volume discounts and accounting treatment. The useful point is directional: the platform makes more when clients transact and use additional tools.
The clean line gets slightly more interesting at OpenPath, The Trade Desk's direct publisher integration. It shortens the route between buyer and publisher, which can sound like supply-side expansion. The company explicitly says OpenPath is neither an SSP nor a yield-management product. Publishers expose inventory; buyers get a more direct path; The Trade Desk maintains that its client remains the buyer. Whether every customer finds that distinction sufficient is a diligence question, but it is a coherent boundary.
“Because we don't own media. We just help you buy it better.”The Trade Desk, platform page
Criteo built a useful loop with two entrances
Criteo describes a broader commerce platform. Commerce Max is a set of demand-side retail media tools. Commerce Yield helps retailers and marketplaces manage and monetize inventory and data. Commerce Grid is explicitly an SSP. In retail media, Criteo says retailers use its technology to monetize audiences, traffic and ad inventory, while brands and agencies use demand tools to buy and measure campaigns.
The money follows those roles. Criteo's 2025 filing says it typically charges retailers a negotiated SSP fee and sometimes a technology fee, while brands pay a negotiated DSP fee. It also says retail media revenue is generally based on a percentage of working media spend and is primarily recognized net because the company acts as an agent in the transaction.
There is a real customer benefit here. Retailers hold purchase data and control high-intent placements near the digital shelf. Connecting that supply to advertiser demand lets a brand buy sponsored products, display or video and then measure sales at product level. Criteo says its data assets gave it exposure to roughly $1 trillion in online sales in 2025. The loop can answer the question advertisers care about: did the impression move merchandise?
The same loop creates the tension. A buyer wants the best available impression at the lowest price consistent with its outcome. A seller wants higher yield and more monetization. A platform paid by both sides can serve both competently, but cannot make those interests identical. If it also controls ranking, access, measurement or fee presentation, governance matters.
The four questions worth stealing
Procurement teams can turn this comparison into a short test. It works for a DSP, retail media network, marketplace or any platform that sits between money and inventory.
- Who pays? List every platform, data, service, auction and supply fee, including rebates or minimums.
- Who supplies? Separate open exchange inventory, direct publisher paths, retailer-controlled placements and any preferred access.
- Who sets the auction? Ask who controls floor prices, ranking, first-price rules, pacing and optimization objectives.
- Who verifies? Determine whether logs, attribution and incrementality testing can be checked outside the selling platform.
For The Trade Desk, the test should probe the full price of data and value-added services, the treatment of supply paths and the freedom to use independent measurement. Buyer alignment does not make every recommendation optimal or every fee cheap. It identifies the principal whose objective the platform says it serves.
For Criteo, the test should add retailer-specific questions. Can a brand see when inventory is available only through a preferred or exclusive integration? Are onsite sponsored-product auctions and offsite buys reported separately? Can the advertiser distinguish the DSP fee from the retailer's SSP or technology economics? What prevents an optimization system from favoring inventory that improves the platform's combined economics over the advertiser's incremental return?
Those questions do not require a buyer to reject a two-sided marketplace. Amazon, exchanges and payment networks all demonstrate why integrated markets can be convenient. The bargain is that convenience must be paired with rules. Clear disclosures, auditable logs, configurable objectives and independent tests turn a potential conflict into a manageable one.
Compare roles before comparing growth
Investors face a related trap. Adtech accounting can make unlike businesses look comparable. The Trade Desk reports revenue net of inventory and supplier components it pays, while its gross-spend metric captures a much larger flow. Criteo's retail media revenue is also primarily net, but its Performance Media products have their own pricing and recognition mechanics. A single revenue multiple can hide differences in media cost, service intensity and the share of transactions recorded as agent rather than principal.
The better comparison starts with a map. Mark the buyer, seller, inventory owner, data owner, auction operator and measurement provider. Then annotate each fee and decide which volumes belong in the denominator. Only after that should margins or take rates enter the conversation.
A practical test can begin with a small campaign rather than a philosophical debate. Hold audience, creative and conversion definition steady. Run comparable cells across inventory sources, preserve impression-level logs where contracts permit, and use a measurement partner that is not paid for selling the media. Look beyond attributed sales to incrementality: which purchases would not have happened without the ad? Retailer data can make attribution feel precise while still crediting purchases that were already likely. Precision and causation are different properties.
Buyers should also resist a false choice between perfect neutrality and useful integration. Every platform has incentives. A percentage-of-spend model rewards more spend, even when the platform serves only buyers. A retailer wants both a pleasant shopping experience and more advertising yield. An agency may negotiate volume economics. The objective is not to find a participant without interests. It is to choose a system in which interests are disclosed, controls are usable and performance survives an outside check.
This is also why the rivalry is larger than a feature contest. The Trade Desk is betting that advertisers will value an independent agent across a fragmented open internet. Criteo is betting that commerce data and retailer integration can produce an efficient market with strong measurement. One sells cleaner alignment. The other sells a tighter loop.
Neither promise removes the need for proof. Independence still needs transparent costs and measurable performance. Integration still needs controls that make competing incentives visible. The contract, auction log and incrementality test are less photogenic than an AI dashboard. They are where trust becomes inspectable.
A product demo shows what the platform can do. The fee model shows what it is encouraged to do.YesPress analysis
Read the filings and product pages
Frequently asked questions
What is the key difference between the companies?
The Trade Desk primarily operates a buyer-side platform. Criteo operates demand-side tools and supply-side monetization products for retailers and media owners.
Does Criteo's structure prove advertiser harm?
No. It creates an incentive tension worth governing and measuring, but the model itself does not prove unfair auctions, steering or harm.
Why choose a two-sided retail media platform?
It can simplify access to retailer inventory and audiences while providing product-level, closed-loop sales measurement across multiple retailers.
Why does The Trade Desk stress independence?
It argues that not owning or operating media lets it prioritize inventory according to the buyer's campaign goals.
What should an advertiser request?
A complete fee schedule, inventory-source reporting, auction rules, data-use terms, optimization controls and independent measurement options.