Confessions of an Attention Fundamentalist
Opinion
Human attention is rich, complex, and multifaceted, and anyone who says otherwise isn’t a fundamentalist.
It’s not every day that you get accused of being a “fundamentalist” in the press, especially when your accuser is a Member of the Most Excellent Order of the British Empire. So, I read Mark Barber’s recent article in these pages with great interest.
Barber lays out an extremely cogent argument to say that our industry’s current obsession with (visual) attention lacks nuance. Understanding attention in general, and attention to advertising in particular, will require more than simply asking “Did they look at it or not?”
To this, I would say: “Preach, brother!” I am in violent agreement with the overall thrust of his article. Human attention is rich, complex, and multifaceted, and anyone who says otherwise isn’t a fundamentalist. They’re an idiot.
But out of respect to Barber’s passion and learning, I think it might be helpful to engage with the specific points he made in the article before asking a new question: Is attention data a map or a model?
Attention is not binary
Barber starts by stating that “attention is a limited resource and can waver from second-to-second depending on how effectively people can focus on the things that matter in the moment and filter out unnecessary distractions.”
Too right. This is why reputable attention providers such as Lumen, Playground.xyz, and Amplified Intelligence measure both the percentage chance of an ad being viewed at all (the binary “on/off” of attention) but also the duration of gaze over time (the “quantum” of attention). The dimension of time is vital in providing a more nuanced view of the phenomenon of attention.
Inattention has value
Sometimes people actively engage with things (including, occasionally, ads!) Other times their eyes glaze over, and they let things wash over them: we can look without seeing. And, on other occasions, we can “take things in” without looking at the screen (c.f. Robert Heath’s Low Attention Processing model). Such variety suggests that a single yardstick of attention (and inattention) is inadequate to describe reality.
Now, I’m not a big believer in Heath’s approach, but even if you are, I am not sure that it proves the point that Mark wants to make. It is, after all, called the Low Attention Model, rather than the No Attention Model. While “covert attention” does exist, it is an acknowledged rarity. Most theorists agree that most of the time, you need some attention in some form to trigger perception, memory, or action. Understanding the nature and impact of these sometimes homeopathic levels of attention is precisely what most attention companies are trying to do.
Attention is not only visual
Sound is important, whether as part of an audio-visual experience or the audio-only variety that you would expect to be dear to the heart of the planning director of Radiocentre. And who could disagree with this? We have ears as well as eyes.
As luck would have it, Lumen has just completed a big project into audio-only media as part of the ongoing dentsu Attention Economy project. The results may help explain precisely why audio-only media such as radio are so powerful in driving results for brands.
But, much as we agree about the value of audio advertising, I would take issue with his description of audio attention. “Put simply,” he claims, “if something is not looked at, then it cannot be seen. However, if something is not listened to, it will still be heard.”
Put simply, this is not true. All attention, even audio attention, is selective. We don’t take in everything we can hear, just as we don’t perceive everything that we can see. In fact, attention scientists observed this selectivity of attention to audio stimulus before they confirmed it with visual media. Donald Broadbent’s famous “filter model of attention” and Colin Cherry’s “cocktail party problem,” both developed in the 1950s, were essentially measures of selective audio attention. Each of our senses, and therefore all of our attention, is finite.
No common currency of attention
Just because all attention is selective does not mean that all attention can be measured in the same way. Eye tracking is a thing; ear tracking isn’t. For the Dentsu study cited above, we had to infer audio attention based on what we know about the relationship between visual attention and memory to create “audio equivalents of visual attention.”
This is a significant challenge and one that I feel most keenly. A lot of assumptions go into the way that Lumen measures and predicts visual attention. Our audio attention measures pile further assumptions on top. How do we know if this tottering heap of assumptions adds up to a hill of beans?
Answering this final question provides a further opportunity for nuance — but also another chance to agree with Barber.
Attention data providers don’t so much provide a map of attention averages as a working model of attention processes. The first step is to collect high-quality attention data and understand its impact on memory and action. But that is only the first step. Then we have to go beyond the basic data to build a model of what drives attention and its impact on memory and action. And then we have to put this model to use in the real world — both to validate the initial assumptions but also to improve and refine their weighting with reference to reality.
When we started Lumen, our initial models were crude and slightly arbitrary. In a sense, they had to be. We had to start from somewhere and “guessed” which factors and categories were likely to relate to attention and memory: size of ad, duration of play, volume of audio, and so on.
But, over time, as we have collected more primary attention data on more media and matched this to more outcomes metrics, we have been able to refine and augment the models. New factors have been included, old factors reweighted, correlations between predictions and outcomes sharpened.
We have gradually knocked some of the edges off our crudest assumptions and incrementally improved the isomorphism between the models and the world. There is still much more work to be done — according to PwC, we only get it right 70% of the time — but the models are built to learn, and learn is what they do.
This process of learning and development through constantly relating everything back to real-world results sounds a lot like the approach advocated by Barber in his article. “Cut out the middleman,” he suggests, “and measure actual advertising outcomes.” For us, the outcomes are an intrinsic part of the inputs — there is no “either/or,” there is only “both.”
This is, I hope you will agree, a nuanced and subtle approach to the topic of attention. There is no room for dogma when you adopt a “learning mentality.”
Mark Barber is right to warn us of the dangers of “attention fundamentalism,” but I’m glad to say he won’t find it here.
Michael Follett is managing director at Lumen Research and one of the media industry’s leading experts on attention measurement and effectiveness. He writes for The Media Leader each month.
Strategy Leaders is our commitment to showcase the media industry’s most important strategy issues and debates, as well as focus on notable ad campaigns and from a media-planning lens.
Sign up for our daily newsletter and receive the Strategy Leaders supplement every Thursday.