Viewability became digital media's quality floor because it could be counted at scale. What it counts is narrow: the conditions under which an ad could be seen.
Whether a person actually looked is a separate question, and Lumen's eye-tracking data puts the answer at only 30% of viewable ads.
That gap now has a common language. The IAB and MRC finalized the first industry attention guidelines in November 2025, and the ARF published live-campaign validation in June 2026.
Attention metrics have moved from vendor pitch to planning input. The open question is who uses them, and for which decisions.
This guide covers what attention metrics measure, how to bring them into a plan where creative and media decisions get made together, how to price attention, and how to validate it before budget moves.
P.S. If you want creative, media, and measurement planned in one room, our integrated marketing services team builds attention into the plan from the brief onward.
One thing before the mechanics. We run creative, media, and measurement as one team because they shape each other, and attention is the clearest proof we know of.
No single function controls an attention score. The platform sets how many seconds are available, the creative decides what those seconds are worth, and measurement decides whether any of it moved the business.
Hand attention data to the media team alone and it gets spent buying longer exposure for ads that were never built to use it. Treat it as a shared input, field-tested like everything else, and it becomes one of the most useful signals in the planning stack.
Attention metrics are measured or modeled estimates of whether a person perceived an ad, for how long, and how deeply. Some are observed directly through eye tracking or panels. Many are modeled from device and placement signals, then projected across a whole campaign.
The cleanest way to place them is as one rung on a measurement ladder:
Each rung answers a narrower question than the one below it, and none substitutes for another. A high attention score with no movement in outcomes is still a failed campaign.
Two distinctions matter for planning. Active attention means eyes on the ad. Passive attention means the ad is on screen or audible while the person's gaze is elsewhere, which is common on TV and audio.
Attention also differs from engagement: a click proves engagement, and the IAB's data-signal explainer warns that even a skip or a mute can reflect attention up to that point. Use attention to judge perception and engagement to judge response, and report them as separate metrics.
Viewability describes the ad's condition, while attention describes a person's response.
Under MRC standards, a display ad counts as viewable when 50% of its pixels are on screen for one continuous second, or two seconds for video.
That threshold says nothing about whether anyone looked. Lumen's 30% figure means most viewable impressions are bought at full price and never seen. In multi-channel plans, that waste compounds quietly, which is why optimizing paid media spend across multiple channels increasingly starts with a quality layer above viewability.
The two metrics depend on each other. The IAB and MRC Attention Measurement Guidelines treat a viewable or audible impression, full sophisticated invalid traffic (SIVT) filtration, and a reasonable indication that a user was present as prerequisites for any compliant attention measure.
Viewability is the gate, and attention is what happens after it. That also settles whether attention metrics will replace viewability: under the IAB and MRC framework, they can't, because every compliant attention measure starts from a viewable impression.
| Planning question | Viewability answers | Attention answers | Use it for |
|---|---|---|---|
| Was the ad technically visible? | Yes | Assumes yes as a prerequisite | Supply eligibility and fraud filtering |
| Did a person look or listen? | No | Yes, observed or modeled | Placement quality comparison |
| For how long? | Time in view only | Attentive seconds | Creative length and branding timing |
| Is this placement worth its CPM? | No | Partly, via cost per attentive second | Pricing and negotiation |
| Did the ad change behavior? | No | No, it predicts but doesn't prove | Pair with lift or incrementality tests |
Action: Keep viewability as a buying floor in every insertion order, and add one attention metric as the quality layer above it.
Attention is measured through five approaches:
They answer different questions, so the useful move is matching each one to the planning decision it can actually inform.
| Method | Typical signals | What it tells you | Main limit | Planning decision it serves |
|---|---|---|---|---|
| Data signals | Time in view, share of screen, scroll speed, audibility, interactions | How exposure conditions vary across placements, at census scale | No single signal measures attention directly | Supply selection and in-flight optimization |
| Visual tracking | Eye tracking, gaze, facial coding, presence detection | Where people looked and for how long | Small consented samples, hardware needed | Creative diagnostics and model calibration |
| Physiological observation | Heart rate, skin conductance, EEG | Emotional and cognitive response | Lab settings; restricted in some countries | Concept and creative pre-testing |
| Panels and surveys | Recall, message association, favorability | Whether the ad registered in memory | Recall decays; sampling bias | Validating that attention became memory |
| Predictive modeling | Data signals calibrated against observed attention | Attention estimates for every impression, including pre-bid | Only as good as its training data | Planning norms and programmatic bidding |
Scoring logic is as important as the method. Some vendors report duration, the predicted attentive seconds per impression. Others, like Adelaide, report a probability of attention tied to outcomes.
Duration-based scores carry a known bias: Adelaide cites Meta research that people over 25 spend 70% longer viewing feed ads than younger users, so optimizing purely to seconds quietly shifts delivery toward older audiences.
Pro Tip: Before adopting a vendor, write down which of the five methods its score rests on and which planning decision you'll use it for. If you can't name the decision, the metric will end up as a dashboard tile.
Attention metrics matter for integrated media planning because each attention score comes from three inputs at once: the platform, the creative, and the audience. A plan built in separate creative and media tracks misprices all three. Let’s dive into the details of that:
Each platform's interface caps how much attention an ad can earn. Karen Nelson-Field's research at Amplified found that 85% of digital ads receive less than 2.5 seconds of active attention, while 89% of out-of-home ads clear that mark. She describes fast-scroll feeds as "inelastic": even exceptional creative can't stretch attention past the limit the platform's design imposes.
For planners, that ceiling is a constraint to design around. A 30-second story placed in a feed with a two-second attention window will be judged on its first two seconds.
Within that ceiling, the creative does most of the commercial work.
Nielsen and NCS analyzed roughly 500 campaigns in 2017 and found creative drove 47% of sales lift from advertising, against 22% for reach and 9% for targeting.
In digital campaigns, creative's share rose to 56%. dentsu's eye-tracking work in 2021 points the same way: the difference between good and poor creative shifted recall by 17%.
The two inputs compound. The Triple Opportunity of Attention study, which combined the Advertising Council Australia's effectiveness database with Amplified attention data and System1 creative testing, found that higher-attention platforms amplify creative's business effects by up to 65%.
Strong creative on a low-attention platform underdelivers, and so does weak creative on a premium one.
This is the claim most attention content avoids: attention data is a creative brief input first and a buying input second.
In most plans we review, attention arrives after the creative is locked, so the media team ends up optimizing exposure for assets that were cut for a different window. The fix is organizational as much as analytical, and it's the same logic behind media and creative handshakes that improve ROAS.
Pro Tip: Put the media planner, the creative lead, and the measurement owner in one attention review before the plan locks, with platform attention norms on the table.
Attention metrics should shape the plan at six points, in this order, with creative and media reading the same data at each step. Each step feeds the next, so skipping one usually breaks the next step.
The table below is the handshake in one view: who does what with the same attention data.
| Planning decision | Attention input | Creative implication | Media implication | Owner |
|---|---|---|---|---|
| Channel roles | Active vs. passive attention by brand size | Decide where the brand needs active attention to build memory | Weight high-active-attention channels for challengers | Strategy lead |
| Placement selection | Historical attention window per placement | Window becomes a brief constraint | Shortlist placements above norm | Media planner |
| Creative edits | Window length and decay | Brand timing and cut length per placement | Map each edit to its placement | Creative lead |
| Plan pricing | Cost per 1,000 attentive seconds | None directly | Compare plans on attention cost and reach together | Media planner |
| Activation | Modeled attention scores | Hold creative constant during supply tests | Attention as a weighted bid modifier | Programmatic lead |
| Learning loop | Creative-ID attention vs. placement baseline | Next brief built from winning edits | Next plan built from winning placements | Measurement owner |
Vendor case studies show what this looks like when it works. In Adelaide's 2025 Outcomes Guide, Haleon's attention-optimized connected TV buying delivered a 133% lift in return on ad spend.
Treat vendor-published results as directional, since the vendor chose which campaigns to feature.
Pro tip: Run the attention review on one shared sheet, with one row per placement and columns for window, edit, cost per 1,000 attentive seconds, and reach. When creative and media argue over the same row, the plan gets better. When they work from separate decks, creative gets blamed for the media's attention ceiling, and vice versa.
Attention CPM is the cost of 1,000 attentive seconds. You get there in two steps: first calculate attentive seconds per 1,000 impressions (APM), then divide the CPM by it.
The comparison below uses illustrative CPMs to show why the cheaper placement is often the more expensive one.
| Placement (illustrative) | CPM | Share viewed | Avg. viewed seconds | APM | Cost per 1,000 attentive seconds |
|---|---|---|---|---|---|
| A: open-exchange display | $8.00 | 35% | 1.5 | 525 | $15.24 |
| B: premium in-article video | $14.00 | 60% | 2.6 | 1,560 | $8.97 |
Placement B costs 75% more per impression and 41% less per attentive second. Published studies apply the same math across media: dentsu and Lumen's 2023 audio research put radio at $0.40 per 1,000 attentive seconds and podcasts at $2.80.
Use the number to decide, with three checks before you act on it:
Our media team reports both CPM and cost per attentive second in every plan review for exactly this reason: either number alone hides the trade-off.
No attention metric is an outcome. Attentive seconds per impression and probability-based attention scores both correlate with recall and conversion, but only controlled tests show that changing attention changed the result.
The evidence is strongest at the channel level and weakest at the single-placement level. The ARF's Phase 3 study tracked four live national campaigns with up to seven independent providers each. It found that television and online video consistently drew more attention than social and display.
It also found that higher-attention channels didn't always produce the strongest brand lift on their own, and that lower-attention channels sometimes delivered real gains when they carried enough media weight.
Vendor data shows correlation at scale. Adelaide's annual Outcomes Guides report average lifts for campaigns optimized to its attention score:
The lower-funnel figure held steady while the upper-funnel average fell as the sample grew. That pattern fits the ARF's finding: attention predicts brand lift less reliably than it predicts placement quality.
Match the metric to the objective, then match the validation method to the claim you need to make:
| Objective | Attention metric to plan on | Creative KPI to pair with it | Validation method |
|---|---|---|---|
| Awareness (challenger brand) | Active attentive seconds per impression | Brand identification rate | Brand lift study with a matched control |
| Awareness (category leader) | Attention-adjusted reach | Distinctive asset presence in the window | Brand tracking plus MMM |
| Consideration | Attentive seconds on longer formats | Message recall | Pre/post panel measurement |
| Conversion | Placement-level attention score as a bid modifier | Conversion rate by creative ID | Geo-lift or incrementality test |
| Supply quality | APM against historical norm by placement | Hold creative constant | Controlled supply test with one vendor |
For the validation column, use each test for a different question:
Our measurement team usually runs the first test on a single channel, holding creative constant, so the result can't be credited to a new edit.
Choose an attention measurement vendor on methodology and per-metric accreditation, never on its headline score. MRC grants accreditation metric by metric and environment by environment, so "accredited" on a sales deck can describe a different product from the one being sold.
DoubleVerify's Authentic Attention metrics show how specific this gets. They were first accredited in January 2023, and in April 2026 the MRC extended accreditation to metrics based on DoubleVerify's authentic non-viewable and modeled methodologies, within desktop, mobile web, and in-app.
Other providers worth evaluating include Adelaide, IAS, Lumen, Amplified, TVision, and Realeyes, each built on a different mix of the five methods.
| Question to ask | Why it matters | Red flag |
|---|---|---|
| What is observed and what is modeled? | Most scaled attention data is predicted | "Proprietary" with no calibration description |
| Which metrics hold MRC accreditation, in which environments, audited when? | Accreditation applies per metric | A company-level "accredited" claim |
| What data trained the model, and how often is it refreshed? | The IAB recommends regular updates to training data | No stated refresh cadence |
| Where does coverage stop? | CTV and walled gardens often need panel data | One score claimed for every channel |
| Can you see the component metrics behind the composite score? | The IAB notes MRC generally requires component reporting | A single index with no breakdown |
| Can the data activate in your DSPs and export to your MMM? | Attention only pays off when it changes a buy | Reporting-only dashboards |
| How is consent handled for person-level data? | Panels and eye tracking require explicit consent | No answer by jurisdiction |
Privacy is one of the more common search questions here, and the answer depends on the method. Data-signal approaches work from placement and device information and don't need cookies or personal identifiers. Eye tracking, webcam measurement, and physiological methods collect person-level data and need explicit consent.
Use the IAB's attention RFI questions to put every vendor through the same checklist.
Attention metrics fail in predictable ways, and five mistakes account for most of them. Each one has a simple guard.
Optimizing to attentive seconds alone can buy attention from the wrong people. Adelaide's "Attentive Audience Paradox" also cites TVision data showing TV ad attention keeps rising with every exposure up to 25, so a duration target can reward heavy frequency on the same viewers.
Guard: Cap frequency inside any attention-optimized line item and check audience composition weekly.
Providers differ in signals, thresholds, training data, and benchmarks, so a score of 60 from one and 60 from another are separate claims.
Guard: Pick one methodology per campaign, and only switch between planning cycles with a parallel-measurement period.
The ARF found measurement approaches agreed at the channel level and diverged at the platform and placement level, which is exactly where buying decisions happen.
Guard: Use channel-level attention for budget splits, and treat placement-level scores as hypotheses to test.
The ARF's 2026 conference summary framed attention as a signal, and stopped short of calling it a trading currency or a standalone predictor of brand lift.
Guard: Negotiate on CPM with attention guarantees as a quality clause, and keep the attention metric out of the billing formula until methods converge.
Shifting budget into the highest-attention inventory can shrink reach below what a brand campaign needs. The ARF's Paul Donato put it plainly: media weight, reach, and frequency all shape whether attention becomes brand impact.
Guard: Set a reach floor before any attention optimization, and revisit the brand and performance budget split if attention targets start eating brand reach.
Attention metrics now have a formal standard. The IAB and MRC published their final Attention Measurement Guidelines on November 12, 2025, after a public comment period that ran through July. More than 200 members of the IAB Attention Task Force helped shape them.
Three other pieces complete the picture:
What hasn't changed is comparability. The guidelines define good practice without forcing vendors onto one method, so scores still differ by provider.
The gaps to watch through 2027 are CTV and walled-garden coverage, where third-party data signals alone remain insufficient.
Attention metrics pay off when the people who read them can act on them together. That means the media planner who knows each placement's window, the creative lead who cuts edits to it, and the measurement owner who tests whether the plan changed outcomes, all working from the same data.
That's how Fieldtrip is built. Our creative, media, and measurement teams sit side by side, so attention norms shape the brief, the plan, and the test design at the same time. We field-test before we scale, and every campaign's attention data feeds the next round of creative.
If you want an attention-informed plan built that way, schedule a meeting with Fieldtrip. The first call covers your current channel mix, where attention data would change a decision, and which test would prove it.
Partly. Non-skippable CTV can inflate time-based attention measures because the ad plays to completion whether anyone watches, so attentive seconds overstate quality there. Plan CTV on presence and probability-based measures, calibrated with panel data, and compare CTV placements only against CTV norms.
Set them per line item. Brand line items get an active-attention target paired with a reach floor, while performance line items use attention only as a bid modifier, judged on cost per acquisition. A single blended attention target across both usually starves the brand lines of reach.
You compare within each platform, then validate across platforms with outcomes. Where third-party tags can't run, providers rely on platform integrations or panel-based calibration, which makes cross-platform scores directional. Use a geo-lift or MMM read to settle cross-platform budget questions, since the attention scores can't.
Enter it at the channel or platform level, as an attention-weighted impression series alongside raw impressions. Placement-level attention is too noisy for most models, and the ARF found methods diverge most at that level.
Run it long enough to reach the sample your outcome metric needs, which for most brand lift studies means several weeks of full-weight delivery. Hold creative constant during the test and decide the threshold for action before it starts.
