How Switzerland measures Radio and TV listener data: an interview with Martin Weber (Mediapulse)
While in Italy, Audiradio takes over radio listening surveys from TER with a CATI research that largely follows what TER did, in Switzerland radio listenership has been measured for years with a “meter” based on Smart Watch. Providing real-time data, as in the case of the collapse in listenership of stations that have just abandoned FM. But are these data significant? FM-world interviewed those who conduct the research, Mediapulse.
This is the report of the interview with Martin Weber, Sr. Research Manager Radio that took place on January 29, 2025. The response regarding the FM shutdown is at the end, but we invite everyone to read each answer: it’s worth it.
Mediapulse
FM-World (Marco Hugo Barsotti): First of all, a few words about Mediapulse and its role in the Swiss media landscape.
Mediapulse (Martin Weber): As an independent industry organisation, Mediapulse is charged with collecting data on the consumption of radio and TV programmes in Switzerland.
This data is considered official currency and is used by broadcasters and programme makers, the advertising industry, government agencies and for research.
As a company with a strong practical focus, Mediapulse seeks to support the Swiss electronic media and advertising market with research services and data products. Modern systems and constant innovations ensure that changes in the consumption of audiovisual media can also be mapped.
Mediapulse stands for a neutral, independent, transparent and scientific approach to media research and is under the supervision of DETEC (OFCOM). A large part of the market is represented on the Board of Directors of Mediapulse AG and the Board of Trustees of the Mediapulse Foundation.
Mediawatch
FM-World: When did you start monitoring radio (and TV) audience? Was the CATI system ever used in the past?
Mediapulse: Radio usage has been systematically measured in Switzerland since 1975.
Until 2000, measurement was conducted through face-to-face interviews (DAR).
Since 2001, radio consumption has been measured via electronic device. The measurement provider has been GfK throughout these years with their portable personal meter based on audiomatching “Mediawatch”. TV measurement began in 1968. Until 1984 with CATI interviews and since then with a TAM system, installed in panel households.
Methodology
MethodologyFM-World: On this web page (https://www.mediapulse.ch/fr/produits/radio-data/) you provide a complete description of your methodology that our readers can consult directly, so let me ask only some details. First, how is the data weighted?
Mediapulse: The data record to be analyzed is weighted daily based on universe criteria established by the Federal Statistical Office. Daily, weekly, monthly, and quarterly weightings are applied. Furthermore, the weighting plan for Switzerland includes the following weighting variables: age groups (5), gender (2), recruitment zones (28), subareas (63), WEMF areas (24), linguistic regions and regional newspaper regions (10).
There’s more: your readers can find more detailed information in a downloadable document on our products page, the document is available HERE.
Mediawatch 4
FM-World: Let’s talk about measurement tools and data provision.
Mediapulse: Radio audience research uses GfK’s “Mediawatch 4” radio research system, composed of a new generation of measurement devices and Sound Sampling Units (SSU).
Regarding EvoRep, there isn’t much more to say than what you’ve found on our website. It’s software provided by GfK. As soon as data processing is completed (with a 7-day delay from the measured day) it’s available through EvoRep for analysis. Every participating radio distributor with a Mediapulse contract has access to all data from all participating radio channels.
Gaussian mon amour
Gaussian mon amour
FM-World: Since “every day at least 1,533 people wear this measurement device,” can we expect your data to have a margin of error (on the entire population) of about 2.5% in the worst case? Is this correct and does it also apply to local radio stations?
Mediapulse: Since measurement is carried out with a rotating sample (with usage periods of 4 weeks or 6 months) the sample size contributing to an analysis depends on the time period and size of the geographic area observed. The sample size for all of Switzerland for a semester is approximately N = 9,000.
The net sample includes panel members who provide actually valid measurement data.
Of the 2019 people in the gross sample, on average every day at least 1533 people must provide data, equal to 75.9% of the total.
From a gender and age group perspective, the net sample structure of people wearing the watch (the device that listens, like those in the cover photo Editor’s Note) is composed as follows:
Variance
Due to the sample structure with many observation days of the same participants that are not statistically independent, we have developed a custom variance estimation suitable for the data.
This estimate is used to calculate the confidence interval for net coverage. The width of the confidence interval depends on the sample size and also on the level of net coverage.
We can give you some examples with anonymous radio stations with different levels of average daily net coverage for linguistic areas (there are practically no analyses made for all of Switzerland because the market is strongly structured in linguistic areas):
Let’s try to decode the table, keeping as benchmark that of Audiradio which can be found HERE.
Mediapulse uses a 95% confidence interval, so if the survey were repeated infinite times then in 95 cases out of 100 the numbers provided will be between “lower” and “upper”.
Let’s take the two Italian-language stations (E and F). For both, data from a sample of 884 individuals are analyzed. We can state that station E has an estimated number of listeners equal to 14.33% of the Italian linguistic area or better, that we are 95% confident that the number of listeners is between 12.74% and 15.92%.
And since the net reach is equal to 14.33% we get a percentage width of the interval relative to coverage equal to 22% (1.38/14.33*100).
The Italian linguistic area in Switzerland consists of approximately 718,000 individuals (this is not indicated in the table, these are our data) so in the example station E has a number of listeners equal to 718,000*.143 = 102,000 (actually, between 91,186 and 114,162).
As for station F, it has a number of listeners between 5,600 and 12,421 (with a lower confidence interval width, magic of statistics).
Total Audience
FM-world: There is a growing trend to monitor the “total audience” of specific programs and not just “live” (linear) program listeners. So the combination of those who listened on air, on demand and this across all devices. Is this a “currency” that you measure?
Mediapulse: So far, only live radio has been within the scope of radio measurement. In agreement with the Swiss radio market, time-shifted listening (i.e., podcasts) is not planned even for the near future.
Minus 23%, significant data?
FM-World: The website tio.ch reported a 23% drop in radio listeners after the FM shutdown.
However, the article states “According to Miko Marr, research manager at Mediapulse company (which measures radio data): «These short-term figures are not significant». The SSR for its part confirms that the data are not sufficiently reliable from a statistical point of view”.
So we’re not asking for comments on this particular decline, but rather an explanation:
Can you explain why the data are not yet statistically significant and when do you expect them to be?
Mediapulse: Mirko Marr’s statement is not directed at statistical significance, perhaps this was lost in translation.
The statement refers to the fact that values consisting of only 14 days (and not even complete calendar weeks) of measurement do not provide a sufficiently high precision value to base a valid statement of general audience loss.
Mediapulse always recommends considering at least a quarter (or more for smaller analysis regions) for reliable results. This is because measurement is carried out with a rotating sample and therefore sample size accumulates over time.
Conclusions
Before answering the question we all ask ourselves – whether turning off FM is a good idea – we wanted to analyze the research methodologies used in Switzerland. And – with Swiss precision and seriousness – we were advised to wait before drawing conclusions.
But one thing is certain: in two months (at the end of the first quarter) we will again request access to the data in order to tell you about them, here on FM-world.it.
(M.H.B. for FM-world)
