OpenX and TVision Launch First Supply-Side Attention Targeting for CTV — Moving Eye-Tracking Data Into the Bid Stream

Attention metrics have spent the past two years moving from research insight to campaign diagnostic to targeting signal. On March 11, they took another step: OpenX and TVision launched OpenX Attention Targeting, the first supply-side attention targeting solution built specifically for connected TV. For the first time, attention data is being used to curate and filter CTV inventory before it reaches the bid stream — not after the campaign runs.

The distinction between supply-side and demand-side matters. When xpln.ai integrated attention segments into Index Exchange's SSP in February 2026, buyers could target attention-scored inventory through Deal IDs on the demand side. OpenX's approach flips the model: the SSP itself filters out low-attention inventory before buyers ever see it, making attention a quality floor rather than a targeting overlay.

How It Works

The system combines two data layers. TVision operates a panel of 5,000 US households equipped with two passive, in-home sensor types: audio fingerprinting via microphone arrays that identify what content is playing on the TV, and optical sensors that detect individual viewers in the room and whether their eyes are on the screen — second by second. No images or audio leave the home; all processing happens on-device. For a viewing session to count as "attentive," TVision requires at least 2.5 seconds of measured eyes-on-screen, as TVision CRO Hassan Babajane explained: "it's got to be 2.5 seconds or greater. That is the trigger."

TVision's panel data feeds predictive attention models that score programming and ad environments across the CTV landscape. OpenX then combines these scores with its own real-time bidstream signals — time of day, device type, viewing duration, and content context — to create a composite attention prediction for each available impression.

The critical filter: OpenX excludes any programming that falls below an attention score of 40%. Inventory that doesn't meet this threshold simply isn't offered to buyers, removing the lowest-engagement placements from the available pool entirely.

Activation Through OpenXSelect

The attention targeting capability is available exclusively through OpenXSelect, OpenX's supply-side curation platform launched in July 2025. Buyers can apply attention targeting to any CTV deal curated within OpenXSelect, whether through self-serve or managed service workflows. The attention layer can be combined with audience, content, and contextual targeting parameters with no additional setup.

OpenX's CTV inventory reaches 231 million-plus monthly unique users through its identity graph, and TV by OpenX — pioneered in 2023 — provides 100% glass-on-wall supply through direct publisher relationships, ensuring ads are delivered to actual television screens rather than mobile or desktop devices misidentified as CTV. That glass-on-wall guarantee is particularly relevant given DoubleVerify research showing that over one in three CTV impressions serve when the TV is turned off, contributing to an estimated $1 billion in wasted CTV ad spend annually.

As Erika Loberg, global head of CTV at OpenX, explained to Digiday: "Attention metrics have historically been very post-campaign-focused. You got a report after the campaign ran. That was great, but it didn't really allow you for any in-flight optimizations."

Supply-Side vs. Demand-Side Attention

The distinction between where attention filtering happens in the programmatic chain has real consequences for how campaigns perform.

In the demand-side model — as implemented by xpln.ai and Index Exchange — attention segments are packaged into Deal IDs that buyers can target through their DSP. The buyer actively selects attention-scored inventory. This gives buyers control but requires them to opt in, configure deals, and manage attention as one more targeting parameter alongside audience and context.

In OpenX's supply-side model, the SSP pre-filters inventory before it enters the auction. Low-attention placements are removed upstream, so every impression a buyer evaluates has already passed an attention quality threshold. This is closer to how brand safety and viewability verification work today — as an infrastructure layer rather than a buyer choice.

Hassan Babajane, CRO at TVision, framed the shift: "The next evolution of attention metrics is from pre- and post-campaign to real-time optimization. This partnership allows buyers to use attention metrics to shape campaign performance."

The Attention Infrastructure Stack

This launch adds another layer to the rapidly maturing attention measurement ecosystem. The infrastructure now includes:

The progression from measurement to standards to activation to curation follows the same path that viewability took a decade ago — from industry debate to MRC-accredited requirement to standard infrastructure. Attention is moving through that cycle significantly faster.

What's Still Missing

OpenX declined to disclose the additional fee for attention targeting, and the company plans to eventually add an estimated reach figure to its interface so buyers can see how applying the attention filter affects campaign scale. That tradeoff — quality versus reach — will be the key decision point for media buyers evaluating the product.

The 40% attention threshold is also worth scrutiny. TVision's panel of 5,000 households, while the largest opt-in CTV attention panel in the US, is still a projection model applied to a much larger inventory universe. The accuracy of those projections across different content types, dayparts, and devices will determine whether the filtering produces genuinely higher-performing inventory or simply a different subset of available impressions.

What This Means for Media Buyers

For CTV buyers, the practical implication is straightforward: if you're buying through OpenX, you can now ensure that your ads are only served during programming where viewers are demonstrably paying attention. The 40% floor means the worst-performing environments are excluded by default.

For measurement teams, the deeper question is whether supply-side attention curation produces measurably better outcomes than demand-side attention targeting — and whether both outperform standard CTV buying without attention signals. The data from Adelaide's 2026 Outcomes Guide, showing 33% brand lift and 53% lower-funnel gains from attention-optimized campaigns, suggests the answer is yes. But those results were measured across campaigns using Adelaide's AU metric, not TVision's scoring methodology. Cross-methodology validation remains an open question as the attention ecosystem matures.

The bottom line: attention has moved from a metric you analyze after the campaign to a filter you apply before the impression. That shift — from insight to infrastructure — is the most important development in CTV measurement quality since viewability verification became standard.