Attention metrics: what to ask before adding another KPI

A report introduces an attention score of 72. Another supplier reports 6.4 attentive seconds. A third provides a predicted probability. All three use the word “attention,” but the numbers are not interchangeable.

An attention metric attempts to describe or estimate aspects of a person’s attention to advertising. The method matters: a score based on delivery signals is different from a direct observation in a research panel.

The IAB and MRC attention measurement framework, finalized in November 2025, covers several approaches, including data signals, visual and audio tracking, physiological observation, and panels or surveys. The accompanying industry material treats attention as a complement to other measures. It does not make every vendor’s score the same unit.

Ask what the number represents

Request a plain-language definition before discussing benchmarks.

Is the result observed or predicted? Does it describe the ad, placement, impression, person, or campaign? If the metric is measured in seconds, what starts and stops the clock? If it is a score, what do the endpoints mean?

Ask which signals are inputs and which are validation outcomes. If viewability is an input to a model, a strong relationship between the score and viewability may partly reflect how the model was built.

The distinction does not make the score useless. It changes what independent evidence is needed to support a business decision.

Check the coverage

A campaign-level average may represent only part of delivery. Ask how many impressions were eligible for measurement, how many produced a usable result, and which environments were excluded.

Consider a fictional campaign with one million impressions and attention measurement on 300,000. The coverage is 30%. If most measured impressions came from desktop while most unmeasured impressions came from mobile, the measured average may not describe the campaign as a whole.

Keep coverage next to the score:

Report field	Example question
Measured volume	How many impressions contributed?
Coverage	What proportion of relevant delivery is represented?
Exclusions	Which browsers, devices, formats, or placements are missing?
Weighting	Is the result adjusted to represent a larger population?
Method version	Did the model or collection method change?

Do not assume the missing impressions would have the same result as the measured ones.

Keep viewability and outcomes in the report

Viewability addresses an opportunity for an ad to be seen under a defined measurement rule. It does not establish that a particular person paid attention. Conversely, an attention estimate does not by itself establish a sale or a causal business effect.

Use the viewability calculator to keep the viewable and measurable impression counts explicit. Then ask what additional question the attention metric helps answer.

For a creative comparison, attention might help identify an execution worth testing further. For a sales decision, it should be evaluated alongside the relevant outcome and study design. A metric that is useful for diagnosis does not automatically belong in billing or automated bidding.

Test whether it improves a decision

Choose one use case before adopting the score across every dashboard. For example, compare two creative treatments in similar placements while keeping audience, dates, and delivery conditions as consistent as possible.

Agree on the business measure and the role of the attention result before seeing the data. Ask whether the score adds useful information beyond the metrics already available. Separate association from causation: higher attention and higher sales occurring together does not show that optimizing the score will cause more sales.

Request validation relevant to the intended channel and task. Evidence from a desktop display panel may not resolve a question about a television interface placement.

Name the metric precisely

Label the provider, unit, method version, coverage, and reporting period. Avoid averaging vendor scores with different scales into a single “attention” total.

Keep the initial adoption reversible. If a new score does not change a decision or explain an issue, it may not need a permanent dashboard column. The goal is a clearer judgment about the campaign, supported by a metric whose meaning the team can explain.