Programmatic Advertising: Why Does TV Measurement Still Rely So Heavily on a Panel in 2026? 🤔

By Max Legrand · GreyFox Ads · A French perspective on TV and digital advertising

Imagine switching on your television and seeing the audience for every channel in real time, right there on your TV's home screen. A media buyer's dream? 🤩 The funny thing is… French viewers could already glimpse something like that back in 2008.

Imagine a different TV audience universe 🌍📺

Picture a parallel universe where switching on the television reveals the live audience figures for each channel on the home-screen mosaic. Imagine what a television buyer at an agency could do with that information. No more waiting to discover which programme attracted viewers: the numbers would be right there.

Except… it is not entirely science fiction. Or rather, it wasn't.

Back to 2008: remember the N9UFBox? 🕚

Let's rewind. The subprime crisis was shaking the world. Bienvenue chez les Ch'tis (Welcome to the Sticks) was breaking French box-office records. Coldplay was everywhere.

And if you were watching TV through a N9UFBox, you could see real-time channel audience indicators on its home screen — for viewers using that service. Those of us over 30 may remember. 😉

That was not a national, person-level TV audience census. It was a glimpse of what an operator could observe across its own connected customer base. Still, the idea was already there.

In 2026, why aren't connected TVs enough? 🤔

Fast-forward to today. The figure cited in the original post is 84% of French televisions connected to the internet, according to Arcom (a reported figure whose precise date and definition should be checked before treating it as a 2026 census). Many connections run through operators' set-top boxes or connected-TV platforms.

So why not simply aggregate the viewing data and calculate audience figures from all those devices?

Because a connected device is not the same thing as a known viewer. A TV can be on without anyone watching. Several people may be in the room. A set-top box can report a channel without telling us reliably who was exposed to it. And operator coverage is not necessarily representative of every French household.

What does the Médiamétrie panel add?

France's television audience measurement has traditionally relied on a representative panel of more than 4,000 households, as described in the original post. It is designed to reflect socio-demographic characteristics of the French population, rather than simply count devices.

Since 2024, the measurement approach has also incorporated aggregated viewing information from certain operators (the precise coverage and methodology can evolve). This is an important distinction: the choice is not necessarily panel or big data. It can be a hybrid of both.

Panels are imperfect. But they provide a way to estimate who is watching, not merely which screen or box is active.

Five reasons people defend panels — and the questions they raise

Some arguments for panels deserve scrutiny. Knowing whether someone is in front of the TV is not the same as measuring their actual attention. And digital services already measure many forms of catch-up consumption. But those observations do not automatically make every connected-device signal comparable or representative.

On GDPR, the claim that every household must always give the same form of consent for every audience-measurement use is too simplistic (the legal basis and applicable ePrivacy rules depend on the specific processing). That does not mean television operators can freely repurpose identifiable viewing histories.

The real opportunity: panels plus large-scale data

Perhaps the better question is not: why do panels still exist? It is: why can't we make better use of the data already generated by connected television?

A well-designed hybrid model could use panels for calibration and demographic representation, and operator or CTV signals for scale and granularity. It would still need to account for non-covered households, shared screens, inconsistent definitions and privacy constraints.

For a French market navigating the growth of CTV and mobile video, measurement quality matters as much as buying technology.

Why do advertisers accept panels on TV but demand individual sales attribution online? ⚖️

And here is the question that really bothers me.

Advertisers can invest millions of euros in television using audience measurement that still depends heavily on representative panels and socio-demographic estimates. Yet those same advertisers sometimes expect digital campaigns to demonstrate sales impact down to the individual user.

That is quite a contrast — particularly when the digital ecosystem must also contend with cookieless browsers, consentless environments and fragmented identities.

Of course, the two channels do not have identical measurement capabilities or objectives. But perhaps the industry should be more consistent about what constitutes sufficient evidence of effectiveness, and more realistic about the limits of individual-level attribution.

For further reading, see our analysis of attribution models in programmatic advertising and the Open Web's attention problem.

So here's the question: why is probabilistic measurement acceptable for television, while digital is so often asked to deliver deterministic proof? We'll leave that one open. 😉

Based on the original LinkedIn post by Max Legrand. The historical figures and references reflect the author's original framing; this page distinguishes panel estimates, device-level observations and legal considerations.