What Is Real Attention Actually Worth And What Are We Paying For It?

Has performance media buying reached its limit?

Netflix closed the second quarter of 2026 with double-digit revenue growth, and the stock fell anyway, on softer-than-expected guidance. In the same period, Disney+ is reportedly discussing a free, ad-supported tier with no subscription at all. These are different stories, but they point to the same instinct: everyone, right now, is trying to widen the audience. The fastest way to do that is to cut price — streaming CPMs, according to eMarketer’s estimates, have fallen consistently over the past three years, and now sit well below where they launched.

There’s a question this scale race tends to leave in the background, and it might be worth bringing back to the center: that wider audience how much of it is actually watching? And how much are we paying, proportionally, not for declared reach, but for the real time people spend with what we buy?

It’s not a question with a clean, definitive answer, but it can be approached with concrete data. The UK — probably the most transparent advertising market in the world — is, for the first time since 2024, a place where you can divide ad revenue by measured viewing minutes, separately for linear TV and for streamers’ ad-supported tiers. The result is an anomaly that gets surprisingly little airtime: broadcaster TV generates roughly £1.50 per thousand minutes of viewing; the same thousand minutes on streamers’ ad tiers are worth roughly £0.85 almost half. (That figure rests on an estimate of ad-tier viewing minutes, based on subscriber share disclosed by the platforms the one input in this comparison that isn’t directly measured.) While TV’s ad revenue is essentially flat, streamers’ is growing at double-digit rates every year.

This is where it gets interesting for anyone watching streamers closely. Their strategy, at this stage, looks clear: win advertising market share by taking ground from linear TV, pushing on volume and low price. The price they charge — and pay — is worth, per minute, almost half of TV’s. What’s missing is that nobody has yet run, for streamers, the equivalent of the exercise coming up shortly for TV, YouTube and social: measuring real attention against price paid, format by format. Until that calculation exists, the question stays open — whether the volume they’re building today is the solid foundation they think it is, or just the latest version of the same trade: more impressions, not necessarily more value.

None of this proves one model is wrong and the other right. It’s a signal that price, right now, is moving faster than the attention that should justify it and that gap almost never makes it into the reports that circulate.

A similar exercise, run not across the whole market but format by format, comes from another angle entirely and it isn’t even recent. Back in 2021, a UK panel had already used eye-tracking, applied to the average CPMs of a single UK client, to measure not just whether an ad was shown, but for how many seconds people actually looked at it. The cost per thousand seconds of observed attention, for the main ad formats, produced a list worth reading in full:

• TV, 30-second spot: £1.70

• YouTube mobile, non-skippable: £1.98

• YouTube mobile, bumper: £2.67

• Instagram, story: £2.90

• Instagram, feed: £5.49

• Facebook, feed: £8.34

• Display, banner: £13.02

Here too, the point isn’t to crown a winning format — each serves different purposes, and context matters as much as format itself. But looking at the list, it’s hard not to wonder whether the price we pay for each of these actually reflects that gap or whether we’ve simply never asked the question closely enough.

Maybe there’s a simple reason this gap stays under-discussed: the way we habitually measure advertising — impressions, reach, CPM — is built to answer short-term questions, not this one. And whoever is deciding a budget, under quarterly pressure, has few tools and even less time to ask the different question.

There’s still a piece of this that depends entirely on the buyer, not on the market at large: generative AI now makes it possible to build different creative for every environment, at a cost that would have been prohibitive not long ago. Using it to produce more, faster, in the same format everywhere, is the obvious choice — but probably also the least useful one, if the actual problem is the gap between what gets spent and how much real attention it earns back.

The pieces to do this properly already exist, and are more mature than they might seem. Contextual DCO for CTV is an established category platforms like Adzymic and Hunch already adjust creative to content genre, daypart, weather and other real-time signals, and Equativ’s “Contextual Intelligence” analyzes on-screen mood and narrative beats to place, say, a luxury travel ad inside a high-end destination documentary. None of this is hypothetical; it’s already sold, named, and reasonably widespread. What seems narrower, and still open, is using attention itself not just thematic relevance as the signal that decides how a piece of creative should be built for a given moment: a fuller, more narrative treatment for a high-attention slot, a two-second hook for a low-attention one, chosen because of what the moment is worth in attention terms, not only because of what it’s about.

A shadow metric for the media plan

If this gap is real, the question becomes practical fairly quickly: what can a CMO actually do with it, without waiting for the whole industry to agree on a new currency?

One option, floated in effectiveness circles more often than it’s actually used: track something like aCPM as a shadow metric a private, in-house calculation run alongside the standard CPM-based reporting an agency already delivers, not instead of it. Nobody needs to renegotiate a contract to start. Take the CPM a plan is already paying by channel and format, and apply, even loosely, the kind of attention hierarchy public research already publishes TV formats near the efficient end, open-web display near the expensive end, video platforms somewhere in between depending on format. What comes out isn’t a new number to report upward. It’s a private lens for asking a sharper question about where a plan’s money is actually going.

What a rough attention audit could look like, using only what already exists

None of this requires new panels, new vendors, or new budget just a different way of reading numbers that are usually already sitting in a plan.

1. Map the current plan against the attention hierarchy. Sort the mix, even crudely, into higher- and lower-attention formats using the kind of ranking above.

2. Apply a shadow aCPM to the largest line items first. Where the current CPM is known, and it usually is, apply published attention ratios to the two or three biggest budget lines, not the whole plan at once. That’s where the exercise pays off fastest.

3. Look for the widest gaps, not the average. The average tells you little. The formats where price and attention diverge most are the ones worth a second look often the ones nobody has questioned in years, simply because they look cheap on paper.

4. Ask the agency for the same view, not a different plan. The point isn’t to walk into a renegotiation with a weapon. It’s to ask a specific question about specific line items, and see how the answer compares with the shadow numbers.

And in Italy, specifically

The UK exercise works because the underlying numbers are already public, assembled once and published for anyone to use. Italy’s version isn’t blocked by a lack of data the pieces are simply scattered rather than assembled. Auditel holds minute-level viewing data by platform. Media agencies and sales houses run attention-adjacent metrics inside their own planning tools, often bought from the same international vendors cited above. The building blocks are already in the building. What’s missing is someone doing the work of putting them together into a single, usable view the same shadow-metric exercise described above, just applied locally instead of imported.

It’s the same question, from another angle, that I’d already tried to raise around qualitative GRP: reach measures the opportunity to be seen, not actually having been seen. In the UK, the numbers to close that gap are already public. In Italy, the data and the tools to try exist too the real difference is knowing how to connect them, not owning them.

Maybe the question worth asking, the next time a cost-per-thousand-impressions figure lands on a planning table, isn’t whether it’s high or low. It’s how much of that price is really buying attention and how much is simply buying its shadow.

Sources: Netflix Q2 2026 shareholder letter; eMarketer, “Digital Video and Forecast Trends Q4 2025”; Thinkbox / Advertising Association-WARC Expenditure Report (2024 data); BARB / Ofcom Media Nations; Thinkbox, aCPM data by format — Ebiquity/Lumen/TVision, “The Cost of Attention Across Media”, 2021 (CPMs used are the average for a single, anonymous UK client; the underlying attention data, by contrast, is a market-wide average across thousands of ads — attention sources: TVision/Lumen UK TV Panel and Lumen digital panels; CPM source: Ebiquity); industry overviews of contextual DCO for CTV (Starti, “Top 15 Dynamic Creative Optimization Platforms for CTV”, 2026; Equativ, “CTV Advertising: What It Is, How It Works, Trends for 2026”).