A Field Guide to Media Quality Metrics: Viewability, Duration & Outcome-Based Attention

Viewability: the floor, not the ceiling

The Media Rating Council (MRC) defines a viewable impression as 50% of pixels being visible for 1 second (display) or 2 seconds (video). That’s it.

No read on quality. No read on engagement. No read on outcomes.

Viewability answered an important question: did the ad have a chance to be seen at all? But that’s where it stops. It treats every viewable impression as equal, regardless of context or quality.

The standard works as a baseline. It doesn’t work as a decision framework.

The MFA problem

Made-for-advertising (MFA) sites make this obvious. These domains deliver ~77% viewability on average, well above the WFA's 63% industry benchmark, while charging CPMs 30–40% below market. High viewability. Low price. Looks great on a dashboard, but doesn’t reliably translate to outcomes.

MFA sites achieve those numbers by flooding pages with ad units, gaming the viewability threshold while delivering almost no value to advertisers.

In practice, optimizing campaigns to metrics like viewability hasn’t improved media quality. If anything, it’s led to more cluttered environments, rewarding inventory that clears the threshold, even when it fails to produce meaningful results.

Duration-based attention metrics: time spent vs. impact

The next evolution of media quality measurement focused on duration: how long someone spends looking at an ad, typically measured or predicted via panel-based eye-tracking.

Duration-based attention metrics (DBAMs) were the first to be consistently tied to outcomes—a relationship demonstrated through early work at Parsec by Adelaide’s founding team.

But duration introduces a different set of problems.

Optimizing for duration

Goodhart’s Law applies quickly: When a metric becomes the target, it stops being a good measure.

Duration reflects a combination of an ad's media environment, the creative itself, and the audience seeing it. Optimizing toward duration doesn’t isolate media quality; it influences all three.

Ad creative drifts toward whatever holds attention—novelty, shock, entertainment—rather than what builds brands or drives outcomes.

The Attentive Audience Paradox

Gaze duration is also heavily shaped by audience demographics.

Older and brand-aware consumers—those more likely to convert—tend to spend more time with ads. Optimizing impressions for longer attention durations can skew delivery toward these groups, at the expense of reaching the audiences most likely to drive incremental business outcomes.

Not all attention seconds are equal

• Platform variability: Three seconds of attention on one platform isn’t equivalent to three seconds on another. Context and environment shape how attention translates into outcomes.‍
• Diminishing returns: The first moments of attention tend to deliver the most value. Chasing longer durations can lead to inefficient spend without proportional impact.

Outcome-based attention: a media quality signal built for results

Outcome-based attention works differently. Instead of treating attention as the goal, it treats attention as one input among many—alongside placement characteristics such as format, size, and position, and environmental factors like device, daypart, and page geometry. These models are trained on full-funnel outcome data to predict a placement’s likelihood of capturing attention and driving business results.

Adelaide’s AU metric brings this approach to life. By isolating the role of media from creative and audience effects, AU provides a consistent signal of quality that advertisers can use to plan, optimize, and buy media more effectively.

Four things separate outcome-based attention metrics:

• Probabilistic, not sampled: AU is a predictive, model-based signal applied at scale, rather than extrapolated from small panels.
• Omnichannel: AU covers ~95% of a typical advertiser’s spend—including display, OLV, CTV, audio, and linear TV—with scores normalized across channels.
• Trained on outcomes: The AU model is continuously trained on full-funnel outcome data. Across 60 case studies in Adelaide’s 2026 Outcomes Guide, AU-optimized campaigns delivered a 33% average lift in upper-funnel KPIs and 53% stronger lower-funnel impact versus other optimization approaches.‍
• Built for activation: AU is available pre-bid and integrated across 125+ DSPs, SSPs, ad networks, and publishers. Leading programmatic platforms like The Trade Desk, Google DV360, PubMatic, and Magnite support a range of activation methods, including AU pre-bid segments, custom bidding, and curated high-AU PMPs.

Three alternatives to outcome-based attention—and where each falls short

Outcome-based attention isn’t the only approach to attention measurement. Three other methods dominate the conversation today. Each addresses part of the problem, but none solves it end to end.

1. Panel-based duration measurement

What it is: An opt-in panel, often webcam-based, directly measuring how long people look at ads. Results get extrapolated to estimate campaign-level attention.

Challenges:

• Limited scale. Panels measure thousands of users; programmatic alone runs on billions of impressions.
• Can’t score every impression pre-bid, so it can’t support real-time buying or optimization.
• Heavily creative and audience dependent. Gaze duration shifts with who’s watching, not just what.

Best for: Post-campaign creative diagnostics. Not a media buying signal.

2. Predicted duration at scale

What it is: A machine-learning model trained on panel-based gaze data, extrapolated across every impression. Solves for scale, but inherits the assumptions of the underlying duration data.

Challenges:

• A model trained on gaze time predicts gaze time. The output doesn’t shift toward business outcomes just because the input is larger.
• Transparency about inputs isn’t transparency about outcomes. Public signal lists may flow to publishers and inventory partners who optimize for the score itself. This is the same dynamic that hollowed out viewability and VCR.
• The harder question for any vendor: has the model been independently audited for its correlation to real business results? For many, the answer is no, or yes, but with weak correlation.

Best for: Creative analysis; directional views of relative attention at scale. Not an optimization KPI tied to full-funnel outcomes.

3. Attention built on verification

What it is: Attention products extended from existing verification stacks, built by the same companies that grew their businesses certifying viewability.

Challenges:

• Inherits viewability-era assumptions. Systems built to confirm an ad was served weren’t designed to evaluate the quality of the environment around it.
• Often relies on engagement proxies (dwell, audibility, interaction) that correlate with surface activity, not business results.

Best for: Teams already standardized on verification reporting who want a marginal lift over baseline viewability.

The common thread

A model trained on duration predicts duration. A model trained on verification predicts verifiability. A model trained on outcomes predicts outcomes. Only one aligns with what brands and agencies are trying to achieve.

How to evaluate an attention measurement partner in 2026

With the Interactive Advertising Bureau (IAB) and MRC releasing Attention Measurement Guidelines, there’s now a clearer standard.

A few questions to ask partners:

• Is the methodology transparent? Published inputs aren’t a substitute for independent validation. Ask any vendor whether a third party has audited their model’s design and its connection to  outcomes, and where the report is published. MediaSense’s review of Adelaide’s AU methodology is one public benchmark.
•  Is your partner independent from verification? Attention and viewability serve different purposes—and should rely on different signals. When attention is built on top of verification, it risks inheriting the same limitations as viewability, rather than offering a more robust measure of media quality.
• Does it cover your full media mix? A fragmented measurement approach leads to a fragmented view of media quality. Eye-tracking works for display and video, but not for audio.
• Is it correlated to business outcomes? Engagement signals are inputs, not results. Look for proof that connects a metric to full-funnel outcomes, like brand awareness, consideration, conversion, and sales.

The bottom line

These metrics aren’t interchangeable. Each provides a different signal and a different level of value:

• Viewability: whether an ad had the opportunity to appear on screen for 1–2 seconds
• Duration: the likelihood that an ad will sustain attention
• Outcome-based attention: a placement’s likelihood to capture attention and drive tangible impact

The first two are worth measuring, but only the third can guide smarter media planning and optimization.

That’s the role of Adelaide AU: a media quality score trained on outcomes, available across 19 channels, and integrated directly into the platforms where media teams make decisions.

Want to understand how your media stacks up on attention and outcomes?