Attention is not a metric.

At Advertising Week this year, I lost count of how many panels included the word "attention" in the title. Attention metrics. Attention economy. Attention as currency. The industry has found its new favorite noun, and it's doing what the industry always does with a new favorite noun: treating it like the answer to a question nobody fully articulated.

The pitch goes like this. Impressions are a bad metric because they only tell you an ad was served, not that anyone saw it. Viewability was supposed to fix that, but it just told you the ad was on screen, not that anyone noticed. Attention metrics go further. They measure whether someone actually looked at your ad, for how long, and with what level of engagement. Progress. Upgrade. Better data. Except the fundamental problem hasn't changed at all.

The proxy problem

Every advertising metric is a proxy. We can't measure what we actually want to know, which is whether the ad changed what someone thinks, feels, or does. So we measure something adjacent and hope it correlates. Impressions were a proxy. Viewability was a proxy. Attention is a proxy. A better proxy, sure. But still a proxy.

The danger is in forgetting that. When the industry talks about attention metrics, the language drifts from "this is a useful signal" to "this is the thing that matters." I've heard media planners say, without irony, that they're now "optimizing for attention." As if attention itself were the objective rather than a waypoint on the road to something else entirely.

I've stared at billboards in traffic for ten seconds at a time. Couldn't tell you what a single one said. Attention without comprehension is just eyeball time. And comprehension without relevance is just cognitive processing that goes nowhere. The metric captures the most visible part of the chain and ignores everything downstream that actually creates value.

Why the industry keeps doing this

There's a pattern here that repeats every few years. The advertising industry identifies a measurement gap, develops a new metric to fill it, and then over-indexes on the new metric until it becomes as meaningless as the one it replaced. We did it with clicks. We did it with engagement. We did it with viewability. And now we're doing it with attention.

The reason is structural. The media supply chain needs tradeable units. Buyers need something to optimize against. Sellers need something to charge a premium for. Attention is perfect for this because it's more sophisticated than what came before, it sounds scientific, and it creates a new axis of differentiation for publishers and platforms that score well on attention metrics. Everyone in the supply chain has an incentive to promote attention as the new standard. The incentive to be honest about its limitations is considerably weaker.

I sat on a panel last quarter with a vendor whose attention measurement product is genuinely impressive from a technical standpoint. Eye tracking, facial coding, scroll behavior, device orientation. The data they collect is extraordinary. When I asked how they validate that higher attention scores translate to actual business outcomes, the answer was a case study from one campaign in one category with one brand. That's not validation. That's an anecdote with a sample size of one.

What attention can and cannot tell you

I want to be fair here because I'm not arguing that attention metrics are useless. They're not. They represent a genuine improvement over what came before, and in certain contexts they're quite valuable.

Attention data is useful for creative optimization. If version A of an ad holds attention for four seconds and version B holds it for one, that's a meaningful signal about creative quality. It's useful for media environment comparison. If the same ad gets twice the attention in one placement versus another, that tells you something about the context. It's useful as one input among many in a media planning process.

What attention data cannot tell you is whether the ad worked. Whether the person who looked at it for four seconds absorbed the message. Whether they'll remember the brand tomorrow. Whether their perception shifted. Whether they're more likely to buy. Those are the outcomes that matter, and attention is at best loosely correlated with them.

I worked with a streaming platform that ran a brand campaign across multiple channels. The attention scores were highest on one particular publisher's pre-roll inventory. Leadership wanted to shift budget toward that publisher based on the attention data. When we ran a brand lift study, the actual brand recall and consideration lift was highest on a completely different channel whose attention scores were mediocre. The ad that people stared at longest wasn't the one that changed their minds.

The measurement we actually need

The honest answer is that the measurement the industry wants doesn't exist yet and may never exist in the real-time, impression-level, programmatically tradeable form that the supply chain demands. The effect of advertising on human beings is slow, cumulative, nonlinear, and deeply contextual. It resists reduction to a single score attached to a single exposure.

The tools that come closest to measuring actual advertising effectiveness are brand tracking studies, marketing mix models, and well-designed incrementality tests. They're slower, more expensive, and less satisfying than a real-time attention score. They also happen to be more honest about what advertising actually does and how long it takes to do it.

None of this is new. Byron Sharp, the Ehrenberg-Bass Institute, Les Binet and Peter Field have been making this argument for years. Advertising works primarily through building and refreshing memory structures that make a brand easier to notice and easier to choose. That process doesn't happen in the three seconds an attention metric captures. It happens over months and years of consistent, broadly-reaching exposure. The metric that matters most is the one you can only see in retrospect.

Better proxies are still proxies

I don't begrudge the industry for wanting better measurement. The desire is reasonable. The money is real and the accountability pressure is intense. If I'm spending fifty million dollars on advertising, I want to know it's working too.

But there's a difference between wanting better measurement and believing you've found it. The attention metrics conversation has crossed from the first into the second, and that's where it gets dangerous. When a proxy gets treated as a ground truth, optimization against that proxy starts to diverge from optimization for actual business outcomes. You end up maximizing attention scores while the brand effects you actually care about remain unmeasured and potentially unmoved.

The most useful thing a marketer can do right now is hold two ideas at the same time. Attention metrics are better than what we had. And they are not what we pretend they are. Use them as a signal. Use them to improve creative and media placement. But don't let anyone tell you that a three-second gaze equals a persuaded customer. Upgrading from a bad proxy to a slightly better proxy is improvement. Calling it transformation is the kind of self-congratulation this industry specializes in.

The measurement problem in advertising didn't get solved this year. It got rebranded.