Ad Budgets Shift From Audience Data to Contextual as Identity Erodes

OpenX has one of the cleaner data sets on this shift: in 2022, 78% of US spend on its platform chased third-party audience segments (the behavioral profiles built from cross-site tracking). By 2025, that was down to roughly 40%. Contextual targeting, placing ads next to relevant content rather than a tracked user, climbed to about 35%, mostly on streaming TV. First-party data spiked in 2024 when Google threatened to kill cookies, then retreated when Google reversed.

The caveat matters. OpenX is heavy in streaming TV, and connected TVs never supported cookies, so contextual was always the default there. More streaming inventory in the mix mechanically deflates the behavioral share without any buyer changing conviction. The open web still runs largely on third-party segments. One exchange's book is not an industry law.

That said, the direction is real even if the magnitude is overstated. Apple's privacy changes and browser restrictions have genuinely weakened identity-based targeting, and streaming is a large, growing channel where contextual is the only option. The structural case for pure-play behavioral data as a premium product is weaker than the 2021 narrative assumed.

Who feels it

For DSPs and agency trading desks, audience PMP deals and DMP seat fees that once looked like competitive edge are starting to look like overhead. The traders running those line items often haven't been told yet.

For data brokers and standalone DMPs (data management platforms), this is the harder story. If behavioral segments have lost nearly 40 points of spend share on a major exchange in three years, the renewal math at current pricing is a problem. LiveRamp (the largest independent identity and data connectivity platform) has a path forward by leaning into clean-room-mediated first-party matching, where two parties reconcile customer records without exchanging raw files. Its ceiling is lower than the 2021 story claimed, but it has somewhere to go. Pure-play third-party segment sellers have less runway.

For mid-tier publishers with no first-party data program, contextual CPMs are rising but still trail behavioral on premium inventory. They're exposed on both sides of the transition.

The call

At least one of LiveRamp, Experian's marketing services arm, or TransUnion's TruAudience will announce a strategic pivot, sale, carve-out, or restructuring of its third-party segment business by Q3 2027 earnings (November 2027). Medium confidence. The direction is clear; which name moves first and when are not. A third-party data business priced on being the identity layer has to re-justify itself when buyers reallocate away from that layer, and the fastest defense of a shrinking standalone asset is to repackage it inside clean rooms or spin it off. Keeping all three businesses positioned exactly as before requires buyer demand to stop eroding. Nothing in the privacy or streaming trend points that way.

Before gutting a data stack, verify contextual performance parity on your own inventory and pull the mix-adjusted number for display-only. Before renewing a DMP seat at last year's price, make the vendor justify it against a share that has roughly halved.

Third-party audience data is losing structural share of ad budgets, and the only hard number we have comes from one SSP. OpenX says that in 2022, 78% of US spend on its platform chased third-party audience segments, the behavioral profiles built by tracking people across sites. By 2025 that was down to about 40%. Contextual, placing ads next to relevant content instead of a tracked user, climbed to roughly 35%, mostly on streaming TV. First-party data jumped in 2024 when Google threatened to kill cookies, then fell back when Google changed its mind.

What's being decided: not a single company's bet, but how the whole buy side and sell side should weight their targeting stack over the next two years. That reallocation is easy to undo in principle (traders reweight templates) but hard to undo in practice once renewal contracts, product roadmaps, and data-vendor seats get rebuilt around it. Nothing sets a hard deadline, but DMP and data-segment renewals are the live clock.

The Market Analyst. The identity graph as a standalone product is deflating, and the money follows. If behavioral segments are going from three-quarters of spend to under half on a major exchange, the pure-play third-party data brokers and DMPs are selling a shrinking asset. That compresses what buyers of those businesses will pay. LiveRamp survives by leaning into clean-room-mediated first-party data, where two parties match customers without handing over raw files, but its ceiling is lower than the 2021 story claimed. The winners are whoever can read and act on contextual signal at streaming-TV scale. In plain terms: owning the data on who someone is matters less; knowing what they're watching matters more.

The Skeptic. One exchange, heavy in streaming TV, is not the US ad market. Cookies never worked on connected TVs, so streaming inventory was always contextual by default. Pile more streaming into OpenX's mix and the behavioral share falls by arithmetic, not by any change in buyer conviction. Third-party segments still run the open web, display, and non-CTV video. And the story undercuts itself: the 2024 first-party spike that vanished the moment Google reversed proves buyers are opportunistic, not converts to a contextual religion. No cross-vertical performance-parity data here. In plain terms: one seller's vivid stat is being presented as an industry law, and the math behind it is a format story.

The Operator. Audience-segment line items are getting defunded while the quarter is still open, and the traders running them haven't been told. The pain lands on audience PMP deals and DMP seat fees that now look like overhead instead of an edge. Streaming campaign setup needs contextual signal built into the workflow, not bolted on at launch. Mid-tier publishers with no first-party data program are exposed: contextual CPMs are rising but still trail behavioral on premium inventory. In plain terms: the targeting the templates default to is quietly becoming the expensive, lower-yield option.

The CFO. A DMP seat and a third-party data subscription are fixed costs defended by a habit, not a return. If 38 points of spend walked away from behavioral segments in three years, the question at renewal is what that seat actually buys now. Contextual doesn't carry a per-segment data tax, which helps gross margin on the buy side. The risk is paying twice during the transition: keeping the old data contracts live while building contextual capability. In plain terms: the data bills that felt like the cost of doing business are starting to look like a subscription nobody re-justified.

The tensions

Is this a trend or a mix effect? The Market Analyst reads a structural repricing of identity. The Skeptic says it's streaming TV flooding one exchange's numbers and nothing more. This is the whole argument. If it's mix, open-web behavioral is fine and the data brokers are safe. If it's structural, their asset is melting.

Move now or wait for parity? The Operator wants DMP seats and audience PMPs renegotiated this quarter. The CFO agrees the costs look stranded. The Skeptic says contextual hasn't proven it matches behavioral performance outside streaming, so cutting the data stack early strands revenue, not cost.

Where's the moat? The Strategist read says signal interpretation (entity extraction, brand-safety scoring at streaming scale) becomes the real edge. The Skeptic says that's a clean story papering over a format-specific reality where display still runs on cookies.

What it hinges on

Two facts decide this. First, how much of OpenX's drop is streaming TV mix versus genuine behavioral decline on the open web. Pull the same split for display-only and you'd know in an afternoon. Second, whether contextual actually performs at parity on premium non-CTV inventory, because that's what unlocks real budget migration rather than a mix artifact.

The council leans toward real but overstated. Identity-based targeting is structurally weaker than the 2021 narrative assumed, Apple and browser privacy changes are real, and streaming is a genuinely large and growing contextual-by-default channel. But "78 to 40" as proof of an industry law is one seller's book. Before anyone guts a data stack, verify parity on your own inventory and get the mix-adjusted number. Before anyone renews a DMP seat at last year's price, make the vendor justify it against a share that's halved.

The Prediction

Prediction: At least one of LiveRamp, Experian's marketing arm, or TransUnion's TruAudience will announce a strategic pivot, sale, carve-out, or restructuring of its third-party segment business by the end of Q3 2027 earnings season in November 2027.

Confidence: Medium. The direction is clear, but timing and which name moves first are not.

Why: The signal is behavioral segments falling from 78% to about 40% of spend on a major exchange in three years while streaming TV, where cookies never worked, takes share. The mechanism is that a third-party data business priced on being the identity layer has to re-justify itself when buyers reallocate away from that layer, and the fastest way to defend a shrinking standalone asset is to repackage it inside clean rooms and first-party matching or spin it off. LiveRamp is already leaning that way. Keeping all three businesses running and positioned exactly as before would require buyer demand to stop eroding, and nothing in the privacy or streaming trend points that way.

Revisit by 2027-11-30: We're right if LiveRamp, Experian, or TransUnion publicly restructures, sells, carves out, or de-emphasizes its third-party audience-segment business by then. We're wrong if all three keep those businesses running and positioned as before.

The streaming shift and the Apple and browser privacy squeeze both pull the same direction, which is why this isn't a bet on one platform's decision. It's a bet that the asset itself keeps deflating.