Attention Metrics Explained: Lumen, Adelaide, and How They Work

TL;DR

Attention metrics measure whether someone actually looked at an ad, not just whether the ad loaded on screen. The two leading platforms are Lumen Research (which uses eye-tracking and calculates Attentive CPM or aCPM) and Adelaide (which assigns an Attention Unit or AU score of 0–100 predicting attention probability and business impact). In November 2025, the IAB and MRC released formal Attention Measurement Guidelines, standardizing how attention should be tracked across the industry. But measurement is not delivery - these tools tell you which placements should perform well, not whether your ad will be seen there. That's the gap Attention as a Service (AaaS) closes: instead of measuring attention after the fact, you buy it upfront as a guaranteed outcome.

‍

Why the industry moved beyond viewability

For most of digital advertising's history, success was measured by impressions - the number of times an ad was served. But an impression only confirms that an ad loaded somewhere on a page. It does not confirm that anyone saw it.

The industry responded by introducing viewability as a standard. The Media Rating Council (MRC) defined a viewable impression as one where at least 50% of the ad's pixels appeared on screen for at least one second (two seconds for video).

This was progress. But it was not enough.

Viewability is a technical metric, not a human metric. It measures whether an ad could have been seen, not whether anyone actually looked at it.

Research by Lumen Research made this gap impossible to ignore: only 30% of viewable ads are actually viewed. The other 70% load on screen and then vanish - scrolled past, ignored, or displayed in a browser tab that no one is actively watching.

This is the invisible ad problem. Advertisers were paying for ads that met the industry's technical standard for visibility but still received zero human attention.

Attention metrics emerged as the solution. Instead of asking "Was the ad on screen?", they ask: "Did someone actually look at it?"

‍

How attention is actually measured: eye-tracking, predictive models, engagement signals

Attention cannot be inferred from delivery logs alone. It requires new measurement technologies that track human behavior, not just technical events.

There are three primary methods for measuring attention in digital advertising:

1. Eye-tracking (direct measurement)

Eye-tracking uses cameras or sensors to monitor where a person's gaze lands and for how long. This is the gold standard for attention measurement - it captures actual viewer focus, not proxies or predictions.

Companies like Lumen Research and Amplified Intelligence use eye-tracking studies to build datasets showing which ad formats, placements, and contexts capture the most attention.

These studies involve real people browsing real websites while their eye movements are tracked. The data reveals:

• Which ads are glanced at versus ignored
• How long viewers spend looking at each ad
• Whether viewers read the ad copy or just scan past it
• Which creative elements (images, headlines, CTAs) draw focus

Eye-tracking is expensive and impractical to deploy at scale across every campaign, so it is primarily used to build predictive models and attention benchmarks that can be applied to broader inventory.

2. Predictive models (scaled measurement)

Predictive models use eye-tracking data as ground truth, then train algorithms to predict attention based on observable signals like:

• Ad format (native, video, display, interstitial)
• Placement (above the fold, in-feed, sidebar)
• Size and position (larger ads in central positions score higher)
• Time-in-view (how long the ad remained on screen)
• Engagement signals (clicks, expansions, video plays)

Companies like Adelaide and Teads use these models to assign attention scores to placements without requiring eye-tracking for every impression.

The advantage: scalability. Predictive models can evaluate billions of impressions in real time.

The limitation: they are still predictions, not direct measurements. A placement that scores well on a predictive model should deliver attention - but there is no guarantee it will.

3. Engagement signals (proxy measurement)

Some platforms measure attention indirectly by tracking user actions that signal focus:

• Video completion rate (did the viewer watch to the end?)
• Scroll depth (how far did the user scroll past the ad?)
• Dwell time (how long did the page stay open?)
• Interaction rate (did the user click, expand, or engage?)

These signals do not confirm that someone looked at the ad, but they provide strong evidence of attention. A viewer who watches a 30-second video to completion almost certainly paid attention to it.

This method is the most accessible - it uses data that already exists in ad servers and analytics platforms. But it is also the weakest proxy for true attention.

‍

Lumen Research: what they do, the aCPM concept, key findings

Lumen Research is the pioneer in attention measurement for digital advertising. Founded in 2015, Lumen has built the largest eye-tracking dataset in the industry, covering desktop, mobile, and connected TV environments.

What Lumen does

Lumen uses eye-tracking technology to measure where people actually look when they browse websites, use apps, or watch video content. Their studies involve thousands of real users in natural browsing conditions - not lab environments.

The data reveals how attention varies by:

• Ad format (video outperforms display; native outperforms banner)
• Placement (in-feed ads receive more attention than sidebar ads)
• Creative (ads with human faces draw more focus than product-only images)
• Context (editorial content retains attention longer than aggregator sites)

Lumen also tracks attention duration - not just whether someone looked, but for how long. This is critical, because research by Dr. Karen Nelson-Field shows that ads need at least 2.5 seconds of active attention to form a memory.

The aCPM (Attentive CPM) concept

Lumen introduced the concept of aCPM (Attentive CPM), which measures the cost per thousand seconds of attention—not the cost per thousand impressions.

The formula: aCPM = CPM ÷ APM

Where:

• CPM = Cost per thousand impressions
• APM = Attentive seconds per thousand impressions (how many seconds of attention 1,000 impressions generate)

For example:

• A $10 CPM placement that generates 2,000 seconds of attention per 1,000 impressions (APM = 2,000) = aCPM of $5 per thousand seconds of attention
• A $25 CPM placement that generates 5,000 seconds of attention per 1,000 impressions (APM = 5,000) = aCPM of $5 per thousand seconds of attention

Both placements cost the same per unit of attention, even though their CPMs are different.

This allows advertisers to compare placements not just on price, but on cost per unit of attention. A cheaper CPM may actually be more expensive per second of attention if the placement captures less focus.

‍

Key Lumen findings

Lumen's research has consistently shown:

• Only 30% of viewable ads are actually viewed (the other 70% receive zero attention)
• Mobile ads receive 40% less attention than desktop ads on average, despite higher engagement rates
• Video ads viewed for 2+ seconds drive significantly higher brand recall than shorter exposures
• Above-the-fold placements receive 2-3x more attention than below-the-fold
• Editorial environments (news, lifestyle publishers) outperform made-for-advertising sites by 50%+ on attention metrics

Lumen's joint study with Ebiquity found a 0.98 correlation between attentive minutes per thousand impressions and incremental profit - meaning attention almost perfectly predicts revenue.

‍

Adelaide: the AU score, how it predicts outcomes, case study results

Adelaide takes a different approach than Lumen. Instead of using eye-tracking, Adelaide built a predictive model that assigns an Attention Unit (AU) score to every placement based on its likelihood of capturing attention and driving business outcomes.

What Adelaide does

Adelaide uses machine learning trained on billions of impressions, eye-tracking data, engagement signals, and downstream conversion data to predict:

1. How much attention a placement is likely to receive
2. How likely that attention is to drive measurable business impact (brand lift, purchase intent, conversions)

The output is a single score from 0 to 100 called an Attention Unit (AU).

• AU 0–25: Low-attention placements (e.g., below-the-fold display banners on aggregator sites)
• AU 25–50: Moderate attention (e.g., mid-page native ads on quality publishers)
• AU 50–75: High attention (e.g., in-feed video ads on premium editorial sites)
• AU 75–100: Very high attention (e.g., full-screen interstitials, pre-roll video on CTV)

How the AU score works

Adelaide's model evaluates dozens of factors, including:

• Placement characteristics (size, position, format, viewability)
• Publisher quality (editorial rigor, audience engagement, brand safety)
• Creative performance (historical data on how similar ads performed)
• User context (device type, time of day, session depth)

The score is calculated in real time for each impression, allowing advertisers to bid higher on high-AU inventory and avoid low-AU placements.

Adelaide's outcomes data: 33% brand lift, 53% lower-funnel impact

Adelaide's 2026 Outcomes Guide analyzed 60 real-world campaigns across 16 industries and found that attention-optimized campaigns (those that prioritized high-AU placements) delivered:

• 33% average lift in upper-funnel brand KPIs (awareness, recall, favorability)
• 53% increase in lower-funnel impact, including conversions and sales

One case study highlighted a financial services brand that shifted 50% of its budget to placements with AU scores above 60. The result:

• 2.6x higher conversion rate compared to campaigns optimized for reach alone
• 40% lower cost per acquisition (CPA) despite paying higher CPMs for premium placements

The takeaway: attention predicts outcomes. Paying more for high-attention inventory delivers better ROI than paying less for low-quality impressions.

‍

The IAB/MRC Attention Measurement Guidelines (November 2025) - what they mean

For years, attention measurement was fragmented. Lumen had its methodology. Adelaide had its AU score. Amplified Intelligence had its neurometric approach. IAS and DoubleVerify had their engagement-based models.

Each vendor defined attention differently, making it impossible to compare results across platforms or create industry-wide benchmarks.

That changed in November 2025, when the Interactive Advertising Bureau (IAB) and Media Rating Council (MRC) released the first comprehensive Attention Measurement Guidelines.

What the guidelines standardize

The IAB/MRC framework establishes:

1. Three tiers of attention measurement

• Tier 1 – Exposure-based: Did the ad appear on screen in a viewable position? (This is baseline viewability.)
• Tier 2 – Engagement-based: Did the viewer interact with the ad? (This includes clicks, video completions, expansions, and time-in-view.)
• Tier 3 – Outcome-based: Did the ad drive a measurable business outcome? (This includes brand lift, purchase intent, and conversions.)

2. Data quality controls

The guidelines mandate:

• Empirical support for predictive models (claims must be backed by real data)
• Methodological transparency (vendors must document how scores are calculated)
• Independent auditing (vendors must submit to MRC accreditation)

3. Cross-media comparability

The framework allows attention to be measured consistently across digital, CTV, social, and out-of-home environments - creating a unified currency for planning and buying.

Why this matters for advertisers

Before the guidelines, buying on attention metrics meant choosing a vendor (Lumen, Adelaide, IAS) and trusting their proprietary methodology.

Now, advertisers can:

• Compare attention across platforms using a common framework
• Benchmark performance against industry standards
• Negotiate with publishers based on guaranteed attention thresholds
• Optimize campaigns using standardized attention data

The guidelines also pave the way for programmatic buying based on attention scores - platforms like DV360 and The Trade Desk are already integrating attention data into their bidding algorithms.

‍

The gap these tools still don't solve: measurement ≠ guaranteed delivery

Lumen, Adelaide, and the IAB/MRC guidelines represent enormous progress. Attention is now measurable, standardized, and tradeable.

But here is the limitation: measurement is not delivery.

These tools tell you which placements are likely to capture attention. They help you optimize campaigns for attention after the fact. But they do not guarantee that your ad will actually be seen.

The problem

Even if you know that a placement scores 85 on Adelaide's AU scale or delivers 3.5 seconds of average attention per Lumen's aCPM metric, you still have to:

1. Win the auction to buy that placement
2. Pay the market rate, which fluctuates based on competition
3. Accept that the score is a prediction, not a guarantee - your specific ad may underperform

You are still bidding on impressions and hoping they turn into attention.

The AaaS difference

This is where Attention as a Service (AaaS) offers a fundamentally different model.

AaaS platforms like VISTY do not sell you placements that should deliver attention. They sell you completed views - verified instances where someone watched your video ad from start to finish.

You are not paying for a predicted AU score. You are paying for an actual outcome.

• No auction: The price is fixed, not bid-based
• No volatility: The CPCV is locked in advance
• No prediction risk: You only pay for completed views that have already been delivered

This is the evolution: from measuring attention → to predicting attention → to guaranteeing attention.

‍

How AaaS closes the gap: from measurement to guaranteed attention

Lumen and Adelaide are measurement companies. They help advertisers understand where attention is likely to occur.

AaaS platforms are delivery companies. They guarantee attention as an outcome.

The comparison

Why both matter

Measurement tools and delivery platforms are not competitors - they are complementary.

• Use Lumen/Adelaide data to understand which placements, formats, and publishers deliver the most attention
• Use AaaS platforms to buy those outcomes directly, without bidding

The ideal workflow:

1. Lumen/Adelaide tells you that premium publisher video placements deliver 4+ seconds of attention on average
2. VISTY sells you 100,000 completed views on The New York Times, Forbes, and Vogue at a fixed $0.05 CPCV
3. You get both: the insight (what works) and the delivery (guaranteed attention)

‍

The bottom line: attention is now measurable, standardized, and tradeable

The attention economy is no longer theoretical. It is operational.

Lumen and Adelaide have built the measurement infrastructure. The IAB and MRC have standardized the framework. And platforms like VISTY are delivering attention as a guaranteed outcome.

For advertisers, the shift is clear: stop paying for impressions that may never be seen. Start buying attention directly.

Discover VISTY Packages →

‍

Frequently Asked Questions

What is the difference between Lumen and Adelaide?

Lumen uses eye-tracking to measure actual viewer focus - where people look and for how long. Adelaide uses a predictive model that assigns an Attention Unit (AU) score from 0–100 based on placement characteristics, engagement signals, and historical outcomes. Lumen provides direct measurement; Adelaide provides scaled prediction. Both are valuable - Lumen for research and benchmarking, Adelaide for real-time optimization.

What is aCPM (Attentive CPM)?

aCPM (Attentive CPM) is a metric introduced by Lumen that measures the cost per thousand seconds of attention. The formula is: aCPM = CPM ÷ APM, where CPM is the cost per thousand impressions and APM is the attentive seconds per thousand impressions. For example, a $10 CPM placement that generates 2,000 seconds of attention per 1,000 impressions has an aCPM of $5 per thousand seconds of attention. This allows advertisers to compare the cost efficiency of attention across different placements and formats.

What is Adelaide's AU score?

AU stands for Attention Unit. It is a 0-100 score that predicts how much attention a placement is likely to receive and how likely that attention is to drive business outcomes. Higher AU scores indicate better attention probability. AU 75+ = very high attention (e.g., CTV pre-roll). AU 25 or below = low attention (e.g., below-the-fold banners on aggregator sites).

What did the IAB/MRC Attention Measurement Guidelines change?

The IAB/MRC guidelines, released in November 2025, standardized how attention should be measured across the industry. Before these guidelines, every vendor had its own methodology. Now there is a common framework with three tiers (exposure, engagement, outcome) and data quality controls. This allows advertisers to compare attention across platforms and creates a unified currency for buying and selling attention.

Can I buy media directly based on attention scores?

Sort of. Some DSPs (like DV360 and The Trade Desk) are integrating attention data from Lumen, Adelaide, and IAS into their bidding algorithms, allowing you to bid higher on high-attention placements. But you are still bidding - pricing is not fixed, and delivery is not guaranteed. For guaranteed attention delivery, you need an outcome-based model like Attention as a Service (AaaS), where you pay a fixed price per completed view.

How much does it cost to use Lumen or Adelaide?

Lumen and Adelaide are measurement platforms, not media buying platforms. They typically charge advertisers and agencies licensing fees for access to their data and tools. Pricing varies based on usage and scale, but expect annual contracts starting at $25,000–$100,000+ for enterprise access. Some DSPs and publishers bundle Adelaide AU scores into their platforms at no additional cost.

Do attention metrics work for non-video campaigns?

Yes, but video is the strongest use case. Lumen and Adelaide can measure attention for display ads, native ads, and other formats, but the signal is weaker than for video. Video completion is a clear, binary outcome (watched to end = attention). Display ads require more complex modeling to infer attention from engagement signals. For non-video campaigns, attention metrics are still useful but less definitive.

Is measuring attention the same as guaranteeing it?

No. Measuring attention tells you which placements should perform well based on historical data and predictions. Guaranteeing attention means you only pay when attention is actually delivered - such as paying per completed view rather than per impression. Lumen and Adelaide measure. AaaS platforms guarantee. Both are valuable, but they serve different purposes.

‍

Last updated: March 2026