xpln.ai and TVision this week announced a strategic partnership that integrates TVision's second-by-second, person-level CTV attention data into xpln.ai's scalable predictive models, aiming to give advertisers a consistent attention measurement framework across connected television, social, YouTube, and the open web. The announcement, dated April 24, 2026, arrives nine days after Viant Technology reached a definitive agreement to acquire TVision for $40 million, a deal expected to close in Q2 2026 that will make TVision's panel data an exclusive capability within the Viant DSP.

The scalability problem in CTV attention

Attention measurement in connected television has long faced a structural tension. Panel-based, person-level data - the most accurate form of attention research - generates granular, verified signals about who is watching, whether their eyes are on screen, and for how long. But panels are inherently limited in size. Translating those observations into actionable signals at the scale of a live programmatic campaign has remained a persistent challenge. Advertisers running campaigns across CTV environments cannot practically wait for post-campaign panel readouts to inform bidding decisions or creative rotation.

According to the announcement, xpln.ai addresses this gap through a unified measurement and modeling platform. TVision's second-by-second attention data serves as a high-quality calibration input within the CTV modeling layer. xpln.ai then combines that data with proprietary datasets, impression-level contextual signals, and additional inputs to produce predictions that extend beyond the panel to the broader CTV ecosystem. The result, according to the companies, is panel-based observations translated into real-time, scalable attention predictions.

TVision captures viewer presence and attention across both linear TV and CTV using passive, in-home computer vision technology deployed across a panel of 5,000 households in the United States. The system operates at the second-by-second level, recording who is in the room, whether their eyes are directed at the screen, whether multiple people are co-viewing, and which program or advertisement is playing. It generates four simultaneous variables - content, delivery method, individuals present, and attention level - that traditional ratings-based measurement has not tracked together.

What the integration does technically

xpln.ai's platform captures 20 to 25 exposure signals per impression and converts them into a single predictive Attention KPI using machine learning models trained on eye-tracking datasets, as described in the company's prior integration with Index Exchange, announced on February 12, 2026. Within that framework, TVision's person-level CTV data becomes the high-quality anchor for the CTV-specific modeling component.

According to the announcement, the partnership enables advertisers to do five things: scale high-quality, person-level attention data across the CTV ecosystem; apply a consistent attention framework across channels and markets; gain visibility into attention in environments where direct measurement is limited; understand which creative and contextual factors drive attention; and optimize campaigns using predictive attention signals rather than delivery metrics alone.

That last point distinguishes this approach from simple post-campaign measurement. Predictive signals can theoretically inform bidding and planning before spend is committed. The xpln.ai platform is designed to feed these predictions into the broader cross-channel measurement framework, so that a buyer running campaigns across CTV, social, YouTube, and the open web can compare attention quality on a consistent basis rather than switching between incompatible measurement systems for each channel.

"To date, high-quality attention data has been constrained by scale and geography, while digital measurement has relied on proxies, leaving advertisers without a consistent way to evaluate performance," said Fabien Magalon, CEO of xpln.ai, according to the press release. "By incorporating TVision's person-level measurement into our CTV modeling, we're extending high-quality attention signals into scalable systems advertisers can apply across their media mix."

Yan Liu, CEO of TVision, added in the same release: "TVision has always focused on measuring real human attention at the person level, capturing what people actually see and engage with on screen. xpln.ai's use of our attention data within their scalable cross-channel models, gives advertisers more consistent and actionable insights across their campaigns."

Context: TVision in transition

TVision is mid-acquisition. Viant Technology announced on April 15, 2026 that it would buy TVision Insights for $40 million - structured as $22.5 million in cash and $17.5 million in Viant Class A common stock at a fixed price established in the merger agreement. Once that deal closes, TVision's panel data and computer vision measurement will operate as an exclusive capability within the Viant DSP, combined with Viant's Household ID covering 95% of U.S. adults 18 and over, and IRIS_ID, the content-level classification system that arrived with Viant's acquisition of IRIS.TV in November 2024.

The xpln.ai partnership, announced while that acquisition is pending, illustrates that TVision's data is continuing to flow through third-party integrations in the period before the transaction closes. Whether those integrations persist after Viant takes ownership is not addressed in the announcement.

TVision, founded in 2014 and headquartered in New York with offices in Boston and Tokyo, has operated broadly as a third-party measurement provider across the industry. Its data has flowed into alternative currency providers, been used in studies by Kargo, and featured in research published by the Video Advertising Bureau in partnership with premium streaming platforms. An OpenX integration announced on March 11, 2026 showed how TVision's panel data could be used to build predictive attention models deployed as pre-bid targeting signals in a supply-side platform. xpln.ai's approach follows similar logic but within a different part of the stack - applying those signals inside a cross-channel measurement and planning framework rather than at the SSP layer.

xpln.ai: background and prior activity

xpln.ai describes itself as a global attention intelligence platform helping advertisers measure, understand, and optimize how attention drives outcomes across YouTube, social, CTV, and the open web. The company combines large-scale eye-tracking data, contextual signals, predictive models, and creative analysis to produce what it characterizes as transparent, research-grade measurement across markets and media environments. Founded in Paris, xpln.ai has its U.S. operations headquartered in New York.

Prior to the TVision partnership, xpln.ai's most visible integration in the programmatic ecosystem was its collaboration with Index Exchange. The February 12, 2026 announcement described how xpln.ai's predictive attention metrics were embedded directly into Index Exchange's supply-side platform, enabling programmatic buyers to activate attention segments - either prioritizing high-attention supply or excluding low-attention environments - before placing bids. That integration was noted in PPC Land's coverage of OpenX and TVision's pre-bid CTV attention targeting launch as evidence that supply-side platforms were increasingly assuming responsibility for operationalizing attention as a real-time optimization lever.

The TVision partnership extends xpln.ai's scope specifically within CTV modeling, where the panel-derived data provides higher-quality calibration than impression-level proxy signals alone could deliver.

Why this matters for advertisers and the broader market

The persistent fragmentation of attention measurement has complicated media planning across channels. In television environments, panel-based measurement from companies like TVision has provided verified, person-level signals. In digital environments including display, social, and open web video, measurement has relied on proxy signals - viewability, time in view, completion rates - that do not directly verify whether a human was actually watching. CTV has sat awkwardly between those two worlds, carrying the inventory characteristics of digital but the viewing context of television.

The VAB and TVision report released in February 2026 illustrated the scale of that gap. Premium video platforms averaged 1 hour and 18 minutes per CTV session against YouTube's 52-minute average - a 49% session length differential that has direct implications for ad frequency planning. Co-viewing stood at 60% for premium video platforms versus 45% for YouTube, a 33% advantage that standard impression metrics do not capture. These figures, generated by TVision's panel methodology, demonstrate both the data's granularity and the difficulty of reproducing that signal at digital programmatic scale.

The Dentsu research cited in the VAB report found that attention metrics improve ROI campaign forecasting accuracy by 38% compared to viewability. Adelaide's 2025 Outcomes Guide found 41% higher brand lift from campaigns using attention metrics. If those figures hold across CTV contexts, the commercial case for applying attention data in planning - rather than relying on impressions and completion rates - is measurable.

What makes the xpln.ai and TVision partnership technically distinct from straightforward measurement partnerships is the modeling layer. Rather than simply licensing TVision data for post-campaign analysis, xpln.ai uses it as a training input for predictive models that operate at scale across CTV environments. According to the announcement, panel-based observations are extended into real-time, scalable attention predictions - meaning the insights generated from 5,000 households are designed to inform decisions across a far larger addressable universe of CTV impressions.

The cross-channel dimension is equally significant for buyers who currently manage separate measurement frameworks for each environment. According to xpln.ai, consistent attention metrics across CTV, social, YouTube, and the open web enable more reliable comparison across channels, deeper understanding of what drives attention, and optimization based on predictive signals. For a media planner building a cross-channel campaign, that comparability matters as much as the absolute quality of any individual signal.

The partnership also sits within a broader industry push toward standardization. Index Exchange's integration with Gracenote in January 2026 addressed content-level transparency in streaming. Nielsen and Adelaide's October 7, 2025 integration combined reach data with attention quality within Nielsen ONE. The pattern across these announcements is consistent: the industry is progressively layering attention signals onto existing measurement and planning infrastructure rather than building separate parallel systems.

xpln.ai is headquartered in New York for U.S. operations. The partnership with TVision, described as delivered through xpln.ai, is available for advertisers using the xpln.ai platform.

Timeline

Summary

Who: xpln.ai, a global attention intelligence platform headquartered in New York (founded in Paris), and TVision, a second-by-second TV and CTV attention measurement company headquartered in New York with offices in Boston and Tokyo, operating a panel of 5,000 U.S. households. TVision is currently subject to a pending $40 million acquisition by Viant Technology, announced April 15, 2026.

What: A strategic partnership integrating TVision's panel-based, person-level CTV attention data into xpln.ai's scalable predictive modeling platform, enabling advertisers to apply consistent attention metrics across CTV, social, YouTube, and the open web using predictive signals derived from second-by-second eye-tracking observations.

When: Announced April 24, 2026. The partnership is delivered through xpln.ai and available now to advertisers using the platform.

Where: Both companies are headquartered in New York. The integration operates across CTV environments globally, connecting into xpln.ai's broader cross-channel measurement framework that covers international markets.

Why: Panel-based CTV attention data has historically lacked scalability, limiting its application in live campaign planning and optimization. Proxy signals used in digital measurement do not verify actual viewer attention. The partnership aims to bridge those two worlds by using TVision's verified person-level observations as calibration data for predictive models that operate at digital programmatic scale, enabling advertisers to compare attention quality consistently across channels rather than managing incompatible measurement systems.

Share this article
The link has been copied!