In recent weeks in the United States, streamers, television channels, and advertisers have been suffering from an ailment that Peter Steinberg at Variety has dubbed Gauge Rage. At the heart of this issue is the Nielsen Gauge, a monthly overview of consumption on connected TV screens, broken down between streaming, broadcast (traditional TV), and cable (pay channels). Here is what the latest Gauge published by Nielsen looked like:

As can be seen, streaming commands the lion’s share with 47% of consumption on connected TV screens in the US, compared to 21.5% for broadcast and 21.2% for cable.

This free metric has often been used by streamers to argue either for their dominance (“YouTube, the leading streaming service in the US!”) or for their relative smallness (“We, Netflix, represent only 8.8% of screen time; that’s less than 10%, it’s low. And even if you add HBO Max, we’ll still be behind YouTube, so let us acquire Warner Bros!” The same argument was used by Paramount). In short, this indicator has gained significant importance within the American ecosystem, to the point of becoming a key reference, bolstered by Nielsen’s reputation as the definitive institute for TV audience measurement.

So what is the problem? Where does this Gauge Rage, currently shaking the American audiovisual sector, come from?

The problem is that Nielsen has decided to incorporate new data (from a study called DASH, conducted by the University of Chicago and the Advertising Research Foundation) into its calculations. The consequence? A short-term expansion in the number of households counted as watching linear television (cable and broadcast).

Naturally, by increasing the number of households counted as watching linear TV channels, this mechanically leads to a relative decrease in the audience share attributed to streaming, causing an unexpected rebound for cable and broadcast.

It is this phenomenon of communicating vessels that has frightened streaming platforms, which fear that the new figures, interpreted without context (which will inevitably happen, let’s not kid ourselves), will devalue their advertising inventories just before the Upfronts, at a time when they are increasingly relying on advertising as a growth driver.

Faced with the outcry, Nielsen first decided to postpone the publication of its next Gauge to March 24, and then finally to April, but without using the new methodology, which it has pushed back to autumn. This is truly a case of taking one step back to take two steps forward.

In statements from its executives, it is clear that Nielsen is somewhat overwhelmed by the significance its Gauge has acquired in the American audiovisual sector. However, the company finds itself even more trapped, caught between its desire to accurately reflect the landscape it studies and the pressure from its clients, who use its data to sell advertising and/or pitch a narrative of continuous growth to investors and Wall Street.

The most virulent official criticism came from the Video Advertising Bureau, which accuses Nielsen of prioritizing the interests of its streaming clients to the detriment of the transparency necessary for the traditional advertising market and the integrity of its data. I have some doubts that the entire US advertising market specifically waits for Nielsen’s Gauge to make its buying decisions, given that they already have access to all the audience data for each individual program, but never mind.

This marks the third methodological upheaval at Nielsen in a short period, following the implementation late last year of its "panel + Big Data" system, which also aimed to better reflect audiences and resulted in a more or less sharp increase in ratings for live broadcast programs. This came just months after the introduction of "out-of-home" viewing metrics, designed to count all viewing occurring outside the home (such as watching sports matches at a bar or at a friend's house).

In all three instances, these implementations tend to boost linear TV program ratings more noticeably than streaming figures (contributing to the recent "records" in sports audiences that seem to indicate a resurgence of interest in these events), though this does not guarantee universal growth. For example, this year's Super Bowl and the Oscars both saw viewership declines compared to the previous year, proving that increases are not automatic across the board.

At the start of 2026, SambaTV also changed its methodology, causing the number of viewer households to sometimes double. Luminate, the third institute I use for my US Streaming Report, has also modified its methodology, a topic I discussed a few weeks ago. Admittedly, regarding Luminate, it seems the changes have become marginal following a rather volatile period, but the fact remains.

The risk with all these methodological changes is the loss of historical comparability. These comparisons are essential for evaluating the performance of various programs, and losing them would be catastrophic. While datasets can always be restarted, these historical benchmarks remain crucial.

I was discussing this with someone recently who said he would bet his life that Netflix would eventually change the methodology for its weekly Top 10, but I am not so sure.

First, Netflix’s Top 10 is not based on panels but on 100% of users worldwide, across all devices. Admittedly, the scope analyzed by Netflix evolves daily depending on the number of subscribers. One could argue that achieving 100 million CVEs (Complete Viewings Equivalent) when Netflix had 150 million subscribers is far better than achieving 100 million CVEs with its current 300 million subscribers. However, this does not change the calculation method itself, which remains perhaps the most direct and basic of all.

Netflix also uses this methodology for its Engagement Reports, and we have almost no blind spots regarding program performance on the platform. Changing this methodology would go against Netflix’s stated desire for transparency (limited, yes, but still far more open than all its competitors combined).

That said, I should note that Netflix has already changed the methodology for its weekly Top 10. Originally, the Top 10 was calculated based on hours viewed before pivoting to CVEs (or "views," defined as total hours viewed divided by the runtime). At the time, this change did not impact me, as I was already using the CVE metric based on the hours viewed announced by Netflix. If Netflix decided to abandon "views" and return to a Top 10 based on hours viewed, it would have virtually no impact on how I cover audience figures.

If I were to sketch out the absolute worst-case scenario Netflix could implement, it would be a shift from hours viewed and "views" to a method based on the number of "viewers," calculated by multiplying "views" by an average number of users per subscription. While this might be more "telling" to the general public, the methodological foundation would be completely shaky. Moreover, here too, we would lose all historical comparability for thousands of programs.

In conclusion, it is worth reiterating that no one exactly knows the true audience figures for any program, whether on linear TV, streaming, or elsewhere. Whether it is Nielsen with its panels, Luminate with its connected TVs, or Netflix with its hours viewed, everyone has significant blind spots that make it impossible to state with precision exactly how many people watched a specific program from start to finish. It is simply impossible. No one knows. No streamer knows how many people are in front of their tablet, phone, or TV watching a program. Nielsen does not know what you are watching on your portable devices, nor does it know what people outside its studied panels are doing. These are surveys. Surveys that aim to be precise, but like all surveys, they may/do not reflect reality. It is not a very selling point for a newsletter focused on streaming audiences, I know, but I have never hidden this fact.