AudienceProject today added attention data to its cross-media campaign studies through a partnership with xpln.ai, a Paris-founded attention measurement company. The combination produces a figure the two companies call deduplicated attentive reach: the number of unique people a campaign reached across channels who, by xpln.ai's measure, paid attention to it.

The partnership was set out in a press release datelined Paris, 27 October 2026, and published on AudienceProject's blog. According to AudienceProject, the studies are available to advertisers and agencies in nine European markets - the United Kingdom, Germany, France, Italy, Spain, Sweden, Denmark, Norway and Finland - with Canada and Mexico to follow. No date was given for either of the later two.

What the announcement leaves out is almost as notable as what it contains. There are no campaign results, no sample sizes, no pricing and no definition of what counts as an attentive impression. The release describes a product and the reasoning behind it, while the mechanics are largely left unstated.

In Short

Two measurement companies have teamed up so advertisers can see not just how many people saw their ads across different channels, but how many of those people actually paid attention. That matters to brands spending on several channels at once, because an ad shown to someone who is looking elsewhere costs the same as one that is watched. Advertisers and agencies in nine European countries can now order a study that reports this number, though the companies have not published exactly how it is worked out.

Two halves of one number

Each company supplies one half of the calculation, and the halves answer different questions.

AudienceProject's half is deduplication. A campaign that runs on YouTube, a broadcaster's streaming app and a group of news sites will reach some people in all three places. Adding up each platform's own reach figure counts those people more than once. Deduplicated reach strips out the overlap and reports how many unique people were exposed, and how often - the reach and frequency figures media planners use to buy campaigns and judge them afterwards.

xpln.ai's half concerns whether an exposure registered with the person in front of the screen. Its approach is grounded in eye tracking. According to the company description in the release, xpln.ai combines "large-scale eye-tracking, contextual signals, predictive models, and deep creative analysis" to measure attention across YouTube, social platforms, CTV and the open web. It was "founded in Paris and now headquartered in New York for U.S. operations," the same text says.

The release puts the division of labour in one sentence: "AudienceProject measures deduplicated reach and frequency across channels, while xpln.ai measures attentive impressions and attention rates." Put together, the new KPI "shows how many of the people reached engaged with the ad," according to the release. AudienceProject calls the result "a game-changer for understanding campaign performance and informing smarter cross-media planning." The announcement offers no data to support that description.

Fabien Magalon, chief executive and co-founder of xpln.ai, framed the metric as a step beyond counting exposures. "Reach tells advertisers how many people were exposed to their campaign. Attention tells them how many of those people engaged. By combining AudienceProject's deduplicated reach with xpln.ai's attention rates, we are giving advertisers and agencies a quality-driven metric that looks beyond ad exposure and evaluates true audience engagement across channels," he said in the release.

Predicted attention, applied to measured people

That sentence carries the most technical detail in the announcement. It also points to a question the release does not answer.

PPC Land's earlier coverage of xpln.ai explains how its attention figures are produced. When Index Exchange embedded xpln.ai's attention signals in its supply-side platform in February, the company's platform was described as capturing 20 to 25 exposure signals per impression - share of screen and time fully in view among them - and combining them into a single predictive attention KPI using models trained on eye-tracking data. Coverage of xpln.ai's mobile deal with Mobkoi in June set out the same design: panel-based eye-tracking trains the model, and the model then scores impressions at scale without live eye tracking on every campaign.

So the attention half of the new KPI is, by xpln.ai's own account elsewhere, a prediction rather than a direct observation of each viewer. AudienceProject's half, by contrast, rests on panel measurement. When the company published a methodology white paper in April, PPC Land reported that it calculates reach from its own consented, weighted panel rather than from platform-reported figures, with official census statistics used to build a synthetic population that anchors deduplication.

Magalon's phrase - deduplicated reach combined with "attention rates" - suggests the two are joined at the level of a rate. One plausible reading is that xpln.ai supplies an attention rate for each channel or placement and that rate is applied to the unique audience AudienceProject measured there. Another is that each exposure recorded for an AudienceProject panellist is scored individually by xpln.ai's model. The two methods would produce figures with different margins of error, and the release does not say which one is used, or how attention metrics calibrated in one market are carried into another.

"Engaged" and "paid attention"

Émilie Carcassonne, managing director for France and Spain at AudienceProject, set out the company's case in the release:

"Deduplicated reach and frequency are the foundation of how advertisers and agencies plan and optimise campaigns across channels, and that foundation has never been more important in a fragmented media landscape. Our partnership with xpln.ai builds on it. By adding attention data to our independent cross-media measurement, we can show advertisers not only how many people in their target group they reached and how often, but also how many of them actually engaged. That extra layer gives advertisers and agencies the confidence to invest where their campaigns work hardest," she said.

Both executives use "engaged" for what the new layer captures. In digital advertising, engagement more often refers to clicks, likes, shares or video completions - things a user does. What xpln.ai measures, as the release describes it, is attentive impressions and attention rates, derived from eye-tracking-trained models, contextual signals and creative analysis rather than from interactions. AudienceProject's own LinkedIn post on the partnership uses a narrower phrase, "paid attention," which sits closer to what the method describes.

What the LinkedIn post adds

AudienceProject promoted the partnership the same morning in a post on its LinkedIn company page. Much of the text tracks the release. The post goes further on what a study delivers, though, and it introduces cost.

"This means advertisers and agencies can now see not only how many people in their target group they reached and how often, but how many of them paid attention, which channels delivered that attention, and at what cost," according to the post. It lists five things a "cross-media attentive reach study" provides: "See who actually paid attention," "Understand what each channel adds," "Compare channels on cost efficiency," "Act on clear recommendations" and "Build on independent measurement."

Cost does not appear anywhere in the press release. It is not new territory for AudienceProject, however. The sample dashboard in its April white paper already reported cost per 1,000 people reached in the target group and cost per 1,000 incremental people reached, according to PPC Land's coverage. An attention layer would make a cost per attentive person reached the obvious next unit. Neither the post nor the release names that unit, and neither explains how spend by channel enters the calculation.

The post closes by thanking Magalon, Laurène Deecke "and the xpln.ai team for the partnership." It does not give Deecke's role.

Studies, not a dashboard

One structural detail in the release is easy to miss. AudienceProject's boilerplate describes "always-on measurement of omnichannel advertising across markets" through "a scalable SaaS platform." The attentive reach product works differently. According to the release, "the cross-media attentive reach measurement is delivered through dedicated studies to advertisers and agencies."

That places attentive reach outside the continuous, self-serve reporting AudienceProject sells as its core product, at least for now. It also cuts against the company's recent direction: Elevate, its measurement environment for broadcasters and publishers, was opened to media owners in June to give their sales teams repeatable reach figures, including cost per 1,000 people reached in a target group, rather than one-off reports built for each pitch. The release does not say whether an attentive reach study runs while a campaign is live, which would allow budgets to move in flight, or only after it ends. Nor does it say whether attentive reach will eventually appear inside the platform.

"Act on clear recommendations," in the LinkedIn post, hints at advisory output alongside the figures. AudienceProject already has a tool that comments on campaign reports: Elevate 3.0, released on August 27, added an AI assistant called Audrey that presents observations drawn only from the data within a given report. Whether the attentive reach studies draw on Audrey, or on AudienceProject's analysts, is not stated.

Nine markets now, two later

The nine launch markets are not arbitrary. They match the footprint AudienceProject is building for other partners. In October, PPC Land reported that the company's Pinterest measurement would expand in the fourth quarter from the UK and Germany to France, Italy, Spain, Sweden, Denmark, Norway and Finland - the same nine countries.

Canada and Mexico, the two markets "to follow," are places where AudienceProject already operates. Its measurement technology became available in Canada through a partnership with NLogic, a subsidiary of the Canadian audience measurement body Numeris, announced in July 2025. Mexico followed in September 2025 as AudienceProject's first Latin American market. In September 2026, the company said Netflix's ad-supported plan was measurable through it in six markets, Canada among them. The delay for the attention layer in those two countries is therefore not a question of AudienceProject's own presence. So why the wait? The release does not explain it.

Poland is absent from the list entirely. It is one of the 10 markets where Elevate is available, but it is not named among the attentive reach markets, either now or later.

What the announcement leaves open

Several questions that matter to anyone evaluating the numbers go unanswered in both sources.

Threshold. The release does not define an attentive impression. Attention vendors use different cut-offs - a minimum number of seconds of gaze, a share of the ad on screen, or a modelled probability - and the choice moves the result considerably. Nor does it say whether a person counts once towards attentive reach after a single attentive impression, or only after attention accumulates across several. That distinction matters because some recent research treats attention as cumulative: Lumen, Amazon Ads and Cint put the threshold for moving purchase intent at a median of 47 seconds of attention summed across impressions. Attentive reach, as described, counts people rather than seconds.

Channel coverage. xpln.ai's description lists four environments: YouTube, social, CTV and the open web. AudienceProject's measurement is described as omnichannel, and its white paper covered linear TV through external measurement data. The release does not say whether linear television, audio or out-of-home exposures in a campaign receive an attention rate, or how they are treated if they do not.

Market data. Nine markets are live from the start. The release does not say whether xpln.ai's eye-tracking training data was collected in each of them. When xpln.ai partnered with TVision in April, TVision's person-level CTV data, drawn from a panel of 5,000 US households, was described as a calibration anchor for xpln.ai's CTV modelling. Whether a comparable European anchor exists is not stated.

Validation. AudienceProject's reach methodology has been audited by independent parties including PwC and Mediasense, according to its white paper. The release does not say whether the combined attentive reach metric, or xpln.ai's attention model, has been through any audit. The Media Rating Council, working with the IAB, finalised attention measurement guidelines in November 2025. The announcement does not refer to them.

Commercial terms and timing. No price, no launch clients and no schedule for Canada or Mexico appear in either source.

Two companies, two trajectories

AudienceProject, which PPC Land has described as a Copenhagen-based independent cross-media measurement company, has spent 2026 widening its partner list. Netflix measurement, the Pinterest expansion and the opening of Elevate to broadcasters all sit on the same thread: a third party that counts unique people across platforms that each report their own numbers. Its role in Europe's wider measurement debate is visible too. When the Coalition for Innovative Media Measurement and the World Federation of Advertisers launched a review of European cross-media measurement in January, PPC Land noted that Médiamétrie's Cross-Media Video solution in France was being built with AudienceProject.

xpln.ai's 2026 has been about distribution. Its attention signals went into Index Exchange's SSP for pre-bid targeting in February. The TVision partnership in April brought in second-by-second CTV data. By June, Mobkoi and xpln.ai said the attention firm had measured more than 1 billion impressions across its mobile inventory over the previous 12 months, with average attention above 5.5 seconds. The AudienceProject deal is different in kind. It takes xpln.ai's attention scores out of buying and inventory packaging and into independent post-campaign measurement, where the number is meant to judge a media plan rather than to fill it.

Why the marketing community is watching attention

Attention measurement has moved quickly from research decks into the buying process. Google put Adelaide's attention scores into DV360 custom bidding from mid-July 2026. Lumen brought attention measurement to Netflix ads in five European markets in March, and in May Teads gained Lumen's eye-tracking measurement for CTV HomeScreen ads. The scale of the gap these vendors are trying to describe is large. TVision, for example, estimated that only 9 of 200 daily viewing minutes are attentive ad time.

Standardisation has lagged. An IAB Europe panel in July 2025 heard that attention metrics influence programmatic decisions but standardisation remains elusive, with vendors producing composite scores that cannot be compared directly. The MRC and IAB guidelines that followed in November 2025 are guidelines, not a shared metric. Lumen counts seconds, Adelaide scores placements, and xpln.ai predicts attention from exposure signals.

Cross-media reach has its own fragmentation problem. Platforms offer deduplication within their own walls: Google Ads introduced Cross-Media Reach reporting in June 2024, deduplicating reach across YouTube campaigns and, in some regions, linear TV. Independent vendors such as AudienceProject sell the opposite proposition, a count that no single seller controls.

What today's partnership does is put the two debates into one figure. For a planner splitting money across CTV, YouTube, social and the open web, deduplicated attentive reach offers a way to compare channels on how many attentive people each adds, rather than how many impressions or how many people were merely exposed. That could change how channels with very different attention profiles are judged against each other. Whether it does depends on details that neither company has yet published: how attention is defined, how it is attached to measured people, and whether the result holds up under independent audit.

Timeline

Summary

Who: AudienceProject, a Copenhagen-based independent cross-media measurement company, and xpln.ai, an attention measurement company founded in Paris with US operations headquartered in New York. Émilie Carcassonne, managing director for France and Spain at AudienceProject, and Fabien Magalon, chief executive and co-founder of xpln.ai, are quoted.

What: A partnership that combines AudienceProject's deduplicated reach and frequency with xpln.ai's attentive impressions and attention rates to produce a new KPI, deduplicated attentive reach, delivered through dedicated studies. AudienceProject's LinkedIn post adds that the studies compare channels on cost efficiency and include recommendations. Neither source defines an attentive impression, explains how the two data sets are joined, or gives prices or results.

When: Announced today, October 27, 2026, in a press release datelined Paris and in a LinkedIn post by AudienceProject the same morning. No dates were given for Canada or Mexico.

Where: Available now in the United Kingdom, Germany, France, Italy, Spain, Sweden, Denmark, Norway and Finland, with Canada and Mexico to follow.

Why: Advertisers buying across several channels have tended to measure unique reach and attention separately. The partnership puts them into one figure, so that channels can be compared on how many attentive people they add. Its usefulness depends on methodological details that have not been published, in a category where vendors still define attention in different ways.