Not all sparkling wine is champagne — and not all attention is equal

Not all sparkling wine is champagne — and not all attention is equal

If it doesn’t have an eye-gaze in it, it’s not attention.


No-one needs another line about how attention is getting a lot of attention these days. But, were we to track the eyeballs of the sharpest marketers at the biggest brands, we’d surely find that many have been looking long and hard at the possibilities attention technologies present.

This summer, Anheuser-Busch InBev signalled its belief that the media industry is making ‘a directional shift’ towards attention as a key trading metric.

Mars, Diageo, L’Oréal and Reckitt are already among the growing body of brands that sit alongside the brewing giant on the Attention Council, convened in 2020 to develop and promote the measurement of attention as an indicator of audience engagement.

2023 will see the noise around attention continue to grow as publishers and vendors promote their high-attention ad inventory, creative agencies trumpet their attention-grabbing executions and measurement companies push their attentive metrics.

Attention is all around us, and everyone will tell you they know how to grab it, convert it and measure it.

But when all of these different organisations use the word attention, are they all talking about the same thing? Is it visual, or aural? Does it use any sort of human input, or does it rely on proxy metrics? And who is defining all of this? 

Pure proxy attention: a recipe for a headache

Attention is a lot like sparkling wine: it comes in many different varieties; it is created in many places with many different processes; and it covers a huge range, from poor to outstanding.

And while you may find some interesting and pleasing cheaper varieties, only the fizz that comes from the Champagne region gets to be called champagne.

Likewise, at one end of the attention spectrum we have what we call pure proxy-based approaches, which use existing metrics or signals such as viewability or time in view to infer that attention has occurred.

The argument goes something like, ‘well, if this ad was up there on the screen for x seconds then the consumer clearly chose for that to occur, and that surely counts as attention’.

Some of these approaches are quite advanced, combining a number of proxy metrics in clever ways. But here’s the problem: at the end of all that, we still don’t know if the person actually looked at that ad.

If we’re continuing the sparkling wine analogy, this approach is the dreaded $5 spumante — exciting enough at the time and maybe it goes down okay, but it’s likely to leave you hungover and wishing you’d looked to see whether there was anything a bit nicer on offer.

Gaze-based attention is the good stuff

At the other end of the spectrum, we have panel-based eye-tracking approaches.

Provided the accuracy of the eye-tracking tech is where it needs to be (this, by the way, is another area where nuance is critical), then this gaze-based approach is a much more pure and reliable way to know attention has occurred. If this was a sparkling wine — well, it would be champagne.

The challenge, of course, is scale — there is simply no panel big enough to reliably handle all the variety on the internet: ad formats, ad lengths, websites, players, audiences and different contexts all create so many permutations that it’s not feasible to observe them all.

This is why some companies, including ourselves, take data from eye-gaze panels and use it to build predictive models for attention to scale the ‘deterministic’ eye-gaze data out in a ‘probabilistic’ way. And to complicate matters, these predictive models use eye-tracking data combined with (you guessed it) the same types of ingredients that we just called Spumante. 

So what’s up there? Well, both sorts of approaches (pure proxy and eye-gaze based) are trying to claim attention occurred. But only the approach with real eye-gaze data in it can actually prove it. 

In data science terms this is called ground truth. And without that, it’s all just speculation and inference.

Choose attention, but choose carefully

The science of attention is complex, which is one reason why the current generic conversation around attention doesn’t always provide brands with the nuance they need to begin working towards meaningful results.

What’s more, some marketers may be waiting for industry standardisation before using attention metrics, and due to the sheer diversity of attention types and methodologies, that might not be as straightforward as it sounds.

In the meantime, brands who are keen to proceed along the attention path, as we believe all smart brands should, need to be sceptical consumers in their own right, trusting only those providers who use robust gaze-based methodologies.

A lot of companies out there are already sticking the attention label on their products without actually having proven attention is occurring, and many more will follow. That is the equivalent of our regrettable spumante badging itself as champagne.

So look at the key distinctions. If it’s not from Champagne, it’s not champagne. And — to reduce a complicated new field to a fundamental principle — if it doesn’t have an eye-gaze in it, it’s not attention.

Rob Hall is CEO of marketing technology company Playground xyz.

Media Jobs