A media buyer sits down on a Monday to launch a campaign. She has a LiveRamp audience she spent six months building. A contextual partner her strategy team prefers. A custom optimization agent her agency has been developing for a year. And a measurement partner the CMO wants her to use.
Her DSP natively supports only one of them.
The rest is workarounds — tagging gymnastics, API duct tape, a deck that explains why the numbers don’t reconcile. By Thursday, half her week is gone to integration work that has nothing to do with the campaign.
This is what most platforms mean when they say they’re agnostic. You can plug things in. Which is true, in the same way you can plug a lamp into an extension cord into another extension cord.
It’s not what Infillion MediaMath means. MediaMath takes a third path. Here’s what that means architecturally.
Agnostic Is a Design Constraint, Not a Claim
On Infillion MediaMath, agnosticism isn’t a feature we added later. It’s a natural result of how the platform is built. Because MediaMath sits on Infillion’s composable infrastructure, the components of a campaign — audiences, contextual signals, measurement, agents, creative — aren’t limited by legacy constraints. They’re interoperable building blocks. Use our components, bring your own, or both orchestrated together on one platform.
That sounds like a small distinction. It isn’t.
The ANA’s latest Programmatic Transparency Benchmark pegged wasted programmatic spend at $26.8 billion in a single quarter. Not because the industry lacks talent or tools. Because the architecture underlying most campaigns is still opinionated — identifiers with limited visibility, optimization for proxy metrics, biased measurement, and more. You can connect outside tools, but they sit at the edges. The center pulls toward one way of running a campaign, and that Monday-to-Thursday tax is what pulling feels like.
Composability inverts that. There is no forced center. The Infillion Platform™ is designed from the start to be assembled from interoperable parts — and operated on by any intelligence, ours or yours. That’s why we can say Infillion MediaMath is agnostic without the asterisks.
Four places this shows up most clearly today.
1. Your Identity Strategy. Not Ours.
Identity is the most contested layer in advertising right now, and it’s the one most platforms are quietly the least agnostic about. The identifier is almost always the lock-in.
MediaMath treats identity as a component. Not the center of gravity.
If you partner with LiveRamp, we can activate against RampID and its connected ecosystem. If you’ve built your own identity infrastructure and want to use it without surrendering it, bring it. Any of these. Combinations of them. Change your mind next quarter.
Underneath all of that sits Infillion XGraph, our privacy-safe identity spine. 250M+ U.S. individuals mapped deterministically across devices and channels, with interoperability built in across UID2, RampID, ID5, and other standards. With therecent addition of Infillion Catalina — one of the world’s largest sources of deterministic purchase data — the deterministic foundation is about to get even deeper. XGraph isn’t a toll you pay. It’s the agnostic identity layer that makes scaled reach possible without sacrificing precision. You don’t have to choose between flexibility and accuracy.
Whatever identity strategy you’ve chosen to trust, MediaMath can activate against it and measure against it. Without asking you to convert.
2. Context That Fits the Campaign, Not the Platform
Contextual is having its second act, and for good reason. Signal quality is decreasing, and privacy pressure is making contextual targeting a necessity rather than a nice-to-have. But “contextual” isn’t one thing — it’s a category of signals, and different campaigns need different ones.
So we integrate contextual partners the way we integrate everything else: as components you choose between. Peer 39, when content understanding is the priority. DoubleVerify when the strategy calls for measurement-backed contextual.
The right partner depends on the campaign — and the platform lets you make that call without re-architecting anything downstream.
3. Measurement That Answers the Actual Question
Measurement is where agnosticism gets tested. Because a DSP that forces you into a single measurement methodology isn’t giving you an outcome — it’s giving you a narrative.
MediaMath treats measurement like every other layer: as a component you compose.
We have our own when you want them. Infillion Arrival for footfall attribution powered by our first-party location data — who visited, where, for how long. Infillion Uplift for real-time brand lift measured directly within ads through zero-party surveys, no guesswork, no extra setup. Both activate within the same workflow. No extra tags. No separate contract.
We also integrate with trusted third-party measurement partners — whatever framework your CMO, procurement team, or agency has standardized on. Paired with log-level reporting updated three times daily, measurement stops being a post-campaign negotiation and starts being a live signal you optimize against.
The point isn’t which measurement methodology is right. The point is that you decide, and the platform accommodates.
4. Any Agent. Yours or Ours.
The agentic era won’t be won by whichever platform ships the best single agent. It’ll be won by the platforms that let any agent operate on them.
The Infillion Agent Connector is built for exactly that. An Infillion agent. A commercial one — Veylan, Newton Research, Mint, HyperMindz. A proprietary one that your agency has been quietly building. The Connector lets any of them plug directly into MediaMath and start planning, optimizing, and executing. No retrofit. No custom integration work that eats a quarter.
Here’s the part that matters. Your agent can only be as good as the platform it acts on. If the decisioning underneath is a black box, your agent is guessing. MediaMath’s optimization engine is the Infillion Brain where every signal and every decision is visible, from bid to budget.. That’s what makes bring-your-own demand agent real. Intelligence is a utility the platform exposes. Not a product the platform sells.
Better agent next year? You’ll use it next year. Nothing rebuilds.
The Compounding Advantage of a Composable Stack
Here’s the part most procurement conversations miss.
The benefit of architectural agnosticism isn’t obvious in year one. In year one, it means you didn’t have to abandon the partners, agents, or data infrastructure you already trust. You moved to MediaMath and kept the pieces that were working. Fine. Table stakes.
Year three is where it gets interesting.
By year three, the agent frameworks have changed. The identity standards have changed. The contextual signals have changed. The measurement methodologies have changed. On a rigid platform, each of those shifts is a migration — a project, a budget, a Q3 eaten by replatforming. On a composable one, each is a component swap. You adopt. You don’t rebuild.
Catalina brings exactly the kind of components a composable architecture is built to absorb. As we continue to integrate it into the platform, MediaMath advertisers won’t be asked to re-platform to use it — it’ll become another signal available in the same workflow. That’s the thesis, in motion.
That’s the actual compounding advantage. Not that MediaMath is agnostic today. That MediaMath stays agnostic through every industry turn still ahead of it.
What Agnostic Should Have Meant All Along
Most platforms treat agnostic as a marketing claim. A partner slide. A procurement checkbox. A word that gets quieter the closer you get to the architecture.
We treat it as a deliberate design choice. The components aren’t dictated to you. They’re yours to compose — audiences, contextual, measurement, agents, creative. Bring what you’ve built. Bring what you’ve chosen. Bring what comes next.
Intelligence is a utility the platform exposes, not a product the platform sells.
Compose the stack. We’ll orchestrate it.
Frequently Asked Questions
What does “agnostic” actually mean for a DSP? In most advertising platforms, “agnostic” means external partners can be connected at the edges while the underlying architecture still privileges a preferred identity graph, optimization loop, or measurement partner. A genuinely agnostic DSP treats each layer of the campaign — audiences, context, measurement, agents, creative — as an independent component that the advertiser can choose, orchestrate, and swap without rebuilding the stack.
How is a composable DSP different from a traditional DSP? A traditional DSP fuses core campaign functions — audience activation, optimization, measurement — into a single, opinionated system. A composable DSP is assembled from interoperable components connected through open APIs, allowing advertisers to bring their own data, partners, and intelligence. The practical difference shows up when the industry changes: on a traditional DSP, each shift is a migration; on a composable one, it’s a component swap.
Can I bring my own AI agent to a DSP? It depends on the DSP’s architecture. Most require custom integration work because the platform’s decisioning layer is proprietary and closed. An agent-ready DSP exposes its decisioning as an open, API-accessible utility — meaning any compatible agent, whether commercial or proprietary, can plan, optimize, and execute campaigns without a custom build. Transparency matters here too: an agent operating on a black-box platform is guessing at the signals underneath it.
