In the past year, attention metrics have seen growing adoption across the advertising ecosystem. The Attention Council, which I co-founded in 2019 with other leaders in the space, published the results of a survey last year showing a three-fold increase in the number of advertisers buying and measuring with attention metrics, with 60% predicting attention metrics would be a primary KPI in the next 1-2 years. It’s not surprising that brands including Anheuser-Busch, NBA and Mars, and agencies including dentsu, Havas, Mediahub and OMD have kicked off dedicated attention practices.

So what’s to come for attention? I think using attention metrics to update media mix modeling (MMM) is likely the next frontier.

MMM’s limiting factor

MMM is having a moment itself as the shifting privacy landscape erodes the usefulness of attribution. Privacy updates by browsers and platforms and looming regulation have limited advertisers’ ability to track individuals along paths to purchase. Even Facebook, the OG of tracking obscene amounts of attribution data, has leaned into MMM as a solution for measuring advertising efficacy with Robyn, its open source automated MMM tool.

But MMM need to be modernized. Traditional media mix models treat all impressions within a channel as if they have homogenous quality, a growing constraint as the fragmentation of digital media accelerates.

Google published a study with Nielsen last September looking at the variance in placement quality that wasn’t being taken into account by MMM. They found ROAS from individual placements typically treated as uniform by MMM varied by as much as 48%.

Attention metrics can help mitigate the damage to models from fragmentation by adding in a meaningful differentiation between placements. One caveat before we explore: in order to be useful for MMM, the attention metrics deployed should focus on measuring media quality. Some types of attention metrics focus on seconds of attention – aka gaze duration – a methodology that is known to be heavily impacted by audience and creative quality. Both are undesirable in MMM.

There are two main ways attention metrics can improve MMM:

Benefit 1: Model training

Setting up a marketing mix model normally involves counting the number of impressions (or ratings points) in a market over a period of time. The model then builds correlations between media delivery and several other factors with business outcomes. As mentioned above, model performance is limited by a lack of accurate measures of quality among placements.

Attention metrics can be used to quantify the quality of the impressions associated with a placement as media mix models are built. They can be as specific as a placement on a page, show / daypart / pod position, or just a different norm for Facebook and Snapchat placements.

Early testing has shown that attention metrics improve the performance of MMM. At Adelaide, as part of a test last year, we partnered with a national retailer to compare the accuracy of marketing mix models that used attention metrics to those that didn’t. The test found that models using attention metrics fit actual outcomes 3x better than models using viewability and 5x better than models using video completion rate for online video.

Benefit 2: Recommendations

Scenario planning tools use MMM to give recommendations of spend across channels based on a set of goals. As media fragments, the variance of quality in a channel increases and the ability of a set of recommendations to consistently deliver outcomes diminishes.

MMM informed by attention metrics can help scenario planning tools make more precise recommendations, and ensure those recommendations result in consistent outcomes, in two ways.

First, suggested impression volumes per channel can incorporate attention floors. In this case, an attention-informed tool might suggest buying 10 million impressions on mobile web placements that meet a minimum attention metric rating at CPM of $10.

Second, price sensitivity to media quality can be expressed using attention metrics. The recommendation may suggest that a 20% improvement over the floor is worth paying 10% more.

Simple applications and next steps

It turns out that MMM may be the easiest way for marketers to work with attention metrics. To get started, have your attention vendor supply your modeling vendor norms at the most granular level that is useful to the model. Then, test the fit of the new model vs. the one without attention metrics.

Media mix models have historically been blunt instruments, thanks in large part to a lack of differentiation of impressions in a channel. This problem is being exacerbated by the accelerating fragmentation of digital media. Attention metrics provide a unique opportunity to increase fidelity of impression quality, leading to more accurate models and more precise recommendations.