Episode 191: Corey Ferengul on the Newest Start-Up Category -- UMPs

Full analysis

Two things happened in this episode that both point the same direction, even though they look unrelated. A new crop of AI startups is quietly moving budget across Meta, Google, and The Trade Desk without a human ever logging in. And the analysts raising their 2026 ad-spend forecasts say the extra money is flowing to Google, Meta, and Amazon precisely because those platforms let you set it and forget it. Same story, two angles: the buyer is being coaxed to stop touching the dials, and a new software layer wants to own the last dial that's left.

The rest of the episode is the ad-tech news you already half-expected. The Google remedy landed soft. DoubleVerify and IAS flirted and walked. LiveRamp drew bidders. All worth covering. None of it is the interesting part.

The interesting part is the category Aperiam is calling Unified Media Platforms, and whether it's a business or a wish.

The Market Analyst. Corey Ferengul from Aperiam Ventures is naming a category and mapping ~22 companies into it, which is what you do when you want to create a fundraising narrative. That's not a knock. It works. But watch what the money is actually rewarding this cycle. Profound raised $180 million at a $1.8 billion valuation for helping brands rank inside chatbot answers, six or seven months after its last round. FLAM raised $40 million for AI creative. The UMP names Ferengul cites raised $10 million-ish. The capital is chasing the interface to the AI platforms. The plumbing beneath gets none of it. Meanwhile Ari Paparo says Magnite and PubMatic will wait "a lot longer" for any Google-remedy windfall. If you're an SSP, the demand shift you were promised is not the trade this year. The trade is that budget keeps concentrating into three walled gardens, and everyone is building tools to feed them faster. The money is betting on the on-ramps to Google, Meta, and Amazon. The open road is empty.

The Skeptic. Steelman the case against UMPs and it holds up uncomfortably well. What has to be true for this to be a durable business? The walled gardens have to keep their APIs open, keep leaving margin on the table, and keep letting a third party sit between them and the advertiser's money. Ferengul says shutting off APIs "isn't an option" because Meta already took away every dial except creative. That's the part I don't buy. Meta stripped the dials because Meta wants the algorithm making those calls, not a startup's agent. A layer whose entire value is re-inserting control the platform deliberately removed is building against the platform's stated direction. And Ari Paparo flagged that some entrants have "no advertising domain knowledge" and are building platform-first with no differentiation. When a VC coins a category and admits a third of the entrants will hit a wall, believe the second half.

The Operator. Forget the category. What happens Tuesday morning when a real buyer turns one of these on? The SMB and franchise fit is real. A two-person team running 40 McDonald's locations genuinely can hand execution to an agent and never log into Ads Manager again. That's a clean win and it's why those are the first customers. The enterprise story breaks the moment you ask an agency how it gets paid. Eric Franchi and Ari Paparo said it plainly: agencies like the back-end efficiency but can't figure out how to keep their media margin if a UMP does the buying. So the tool that saves the agency work also eats the fee that justifies the agency. Nobody sells the thing that shrinks their own paycheck. Second-order effect at 90 days: the SMB deployments hum along, the enterprise pilots stall in procurement, and the vendors quietly reprice as SaaS seats because spend-based pricing scares off the only customers who could pay real money.

The Customer / End User. From the advertiser's chair, this is genuinely appealing and genuinely dangerous. Appealing because logging into five platforms to reconcile budgets is miserable, and Eric Franchi's point that more ad spend is buying the same eyeballs more efficiently is exactly what a CMO wants to hear. Dangerous because every layer that promises to end walled-garden lock-in is itself a new lock-in. You hand budget authority to an agent that shifts money every five to ten minutes, and now you can't audit a single decision, can't unwind it fast, and can't easily leave the UMP without rebuilding your whole buying operation. Are advertisers asking for this? SMBs, yes, because they have no one to do the work. Large advertisers are asking for control and transparency, and an autonomous agent gives them neither.

The CFO. The economics only work at one of two extremes, and the middle is a graveyard. Serve thousands of SMBs on cheap SaaS seats and you're running a low-margin volume business against Meta's own free tools, which keep getting better at the same job. Serve enterprises on a cut of media spend and you're asking a client to route millions through your agent while you're a startup with $10 million in the bank. Ferengul says Blaze and Omni Key already clear $10 million in revenue, which is real and worth respecting. But revenue at that level tells you the SMB motion works, not that the enterprise motion exists. The payback question is brutal: the platforms these tools depend on are the same platforms with every incentive to absorb the function and charge nothing for it. You are building margin on top of a company that gives the same feature away to win the ad dollar underneath you.

Where the council splits. Ferengul thinks UMPs "upend a lot of players in the Lumascape" by compressing the DSP and pushing spend toward supply. The Skeptic and the CFO think the platforms simply absorb the function, because a middle layer that depends on someone else's API and someone else's margin has no defense when that someone decides to compete. That's the real disagreement, and it's the whole ballgame. Second split: the Operator says the SMB business is genuinely good; the Market Analyst says a genuinely good SMB business is not what's being funded or pitched. Both can be true, and if they are, the category gets repriced hard when the enterprise story fails to show up.

What it hinges on. One belief: do Meta, Google, and Amazon want an autonomous third party sitting between them and the advertiser, or do they want to be that agent themselves? Every forecast in this episode says spend is concentrating into those three because their own automated tools work. That is the platforms telling you exactly what they intend to own. A UMP is betting the platforms will tolerate a competitor for the customer relationship. The 2026 spend numbers are the platforms answering no.

Prediction: By the end of Q1 2027 earnings season, at least one of Meta, Google, or Amazon will ship or expand its own cross-account, AI-driven budget-allocation tool that does natively what Unified Media Platforms sell as a third-party layer, undercutting the category's core pitch.

Confidence: Medium. The platforms' incentive is obvious, but timing on a specific ship is the soft spot.

Why: The episode's own evidence is that Brian Wieser at Madison Avenue Consulting and the IAB both raised 2026 forecasts specifically because Google, Meta, and Amazon's automated buying tools are pulling in incremental spend, and Corey Ferengul concedes Meta already stripped every marketer control except creative to keep the algorithm in charge. A platform that removes dials to own the decision does not then hand that decision to a startup's agent for free; the cheapest way to kill a middle layer is to offer its function natively at no added cost, which the platforms already do inside a single account and have every reason to extend across accounts. The opposite outcome, platforms leaving that budget-shifting margin to third parties indefinitely, contradicts the exact behavior the analysts are crediting for the spend surge.

Revisit by 2027-05-15: We're right if any of Meta, Google, or Amazon announces or materially expands a first-party tool that allocates budget across accounts or campaigns using AI automation in a way that overlaps the UMP pitch. We're wrong if none of the three ships or expands such a capability and the UMP startups named here (Blaze, Omni Key, Valence, Minerva) raise up-rounds on enterprise traction over the same window.

One more thing for the SSP crowd. Ari Paparo's read on the Google remedy is the quieter loss here. Magnite and PubMatic priced in demand walking off AdX, and the remedy left DV360 untouched and gave AdWords a wide ROI carve-out to keep funneling to AdX. The publisher ad-server switching is real and useful, but the demand windfall you modeled is not arriving on the schedule you modeled it. Plan accordingly.