Some radios profited from non-existent listeners

Last fall, the topic of simulcasting suddenly made waves: In mid-September, the radio stations Basilisk and Bern1 publicly sounded the alarm and accused Mediapulse of "inertia" and "cover-up". They also put political pressure on the research foundation: a parliamentary question was put to Media Minister Doris Leuthard. But what was it about? If two or more stations broadcast the same program at the same time - a phenomenon known as simulcasting - the Mediapulse measurement system must assign the measured usage to the correct station. Simulcasting has occurred frequently in the past and was not a problem for short sequences (news, music) - Mediapulse even carried out a review in 2013. The Research Foundation only recognized major differences at the beginning of 2014 after local radio stations repeatedly drew its attention to them. The reason: Ringier radios Energy Zurich and Energy Bern were the first to increasingly expand their simulcasting from around 2012, and Energy Basel was also included from March 2014. Initially, this only happened from time to time, then for two hours a day and since 2014 for four or more hours a day. However, the SRF and RTS radio stations as well as the French-speaking Swiss stations RFJ, RJB and RTN from Pierre Steulet (Groupe BNJ) also frequently broadcast longer program sections synchronously. As a result, the station groups with extensive simulcasting were reported by Mediapulse as having more listeners than they actually had.

System "invents" additional listeners

To understand this, you need to know how radio measurement works at Mediapulse. This is done with the Mediawatch and is based on audio matching. The wristwatch uses an integrated microphone to record and encode the sounds in the test person's surroundings. Mediapulse later compares this data with around 150 domestic and foreign radio stations whose programs it has also recorded and coded. If the codes on the Mediawatch match the code of a program at the same time, it is established that the test person listened to this station during the time in question.

With simulcasting, however, at least two stations have the same code at the same time. Which one can the respondent be assigned to as a listener? With short sequences, this is relatively easy because clear indications before or after (such as trailers, announcements, etc.) allow the correct assignment. However, if several radios broadcast an absolutely identical program for hours on end, even a longer listening sequence does not contain any clear indication of the station tuned in. If the test subject listens to the station with interruptions during this time - for example because he leaves the room from time to time - the system has to find the correct assignment one after the other, but without any clear indication of the station. The following happened: Although Energy Zurich, for example, was switched on, the system also assigned the same listener to the same-sounding Energy Bern and Energy Basel in between. As a result, the stations in Bern and Basel each received one listener and thus a higher reach - and the entire Energy group received a total of three listeners, two more than in reality. Of course, these represent several thousand listeners each due to the extrapolation. Incidentally, it was not quite the same the other way round: Energy Zurich benefited less from its sister stations in Berne and Basel, as the measurement system tends to favor the smaller station at the expense of the larger one.

A solution has been found

However, not all "multiple" listeners of simulcasting stations have to be wrong. If, for example, a driver travels from Zurich to Bern, listens to Energy Zurich at the beginning and then switches to the same-sounding Energy Bern later, a listener must logically be assigned to both stations. Mediapulse's problem now is to distinguish plausible multiple listeners from implausible ones. Mediapulse and the radio industry have now optimized the measurement system so that the data from the second half of 2014 could be corrected retroactively to a certain extent and, above all, similar errors can be avoided in the future - even if simulcasting increases. The solution works as follows: Mediapulse first compares the codes of all 150 recorded stations with each other to determine when which stations were simulcasting. So that listeners can then be assigned to the correct station, the system now "draws on additional plausibility checks and previous usage sequences of the listener".

However, only the figures for Energy radios and Steulet stations could be corrected retrospectively; those for SRF and RTS stations could not be corrected for technical reasons. For this reason, the RTS and SRF stations were not shown individually in the 2014-2 semester publication, but only as a group. The simulcasting effect is canceled out in the group. From the 2015-1 publication onwards, however, the SRF and RTS stations will be shown individually again.

Up to 15 percent too many listeners per station

However, one question remains: How much of an impact did the misattributions in the 2014-1 semester publication (and earlier) have? Opinions differ. Last summer, Mediapulse cited an example in which the Energy stations "gained" a total of 13,000 additional listeners or 12% more reach in the first half of 2014 thanks to simulcasting. In contrast, the radio stations Basilisk and Bern1 spoke of 50,000 to 100,000 double or triple listeners at times in September 2014. They criticized that this had given Energy an undeserved advantage in the advertising market. An approximation based on the 2013-1 to 2014-2 semester figures gives more of an average value and should also indicate the extent for Energy Bern and Basel and their direct competitors. The procedure (see charts): First, the reach trend for all stations is drawn as a curve based on the official figures, including reach 2014-1, which was most clearly too high for Energy radios (we ignore earlier errors at Energy Bern for the sake of simplicity). If we now connect the values for 2013-2 and 2014-2 directly in the curve (without 2014-1), this indicates the lowest possible reach curve without the simulcasting effect. It can be deduced from this that simulcasting provided the stations Energy Basel and Bern with 11% and 14% additional reach per day respectively in the first half of 2014, and thus a significant increase of 26,800 listeners in total. However, it should be noted that this approach does not take into account other factors such as programming or listener behavior, which also influence the development of reach: The simulcasting effect is therefore more likely to be smaller. But precise figures are not possible.

The industry breathes a sigh of relief

But no matter how strong the simulcasting effect was - what do the radios, especially Energy competitors Bern1 and Basilisk, have to say about the solution? "I am satisfied with the current state of affairs and am confident that we will not experience any surprises here," writes Peter Scheurer, Managing Director of Radio Bern1. The errors have been localized and intercepted with new filter rules. "The audience corrections, such as those made at the Energy stations, make sense. Unusual effects, such as excessively high hourly reach during simulcasting phases, can no longer be identified." Matthias Hagemann, owner of Radio Basilisk, takes a similar view, although he criticizes Mediapulse's behaviour. For a long time, the company did not really help to clarify the matter. This only changed when Franziska von Weissenfluh took over the chairmanship of the Board of Directors and Foundation Board.

The private radio association VSP and the SRG also agree with the solution that has been found. It is important that trust in the radio currency is maintained, they say in unison. This view was also shared by the Interessengemeinschaft Elektronische Medien (IGEM).

Markus Knöpfli

"Nobody is perfect": other research systems also have flaws

Mediapulse's radio research is not the only media research system to be (temporarily) flawed. The best-known example of a measurement error was provided two years ago by Mediapulse's new TV measurement system: here, too, there were initial allocation mishaps, which led to legal wrangling and a data blockade lasting a good six months. But it didn't stop there: in April 2014, the media science commission, which regularly reviews the panel, found a few more "intolerable" flaws, which have since been "significantly" improved. However, another small but embarrassing error recently emerged: Kantar Media, the operator of the TV panel, had "discovered a misadjustment" at the end of January 2015: The panel, which measures the TV usage of people aged 3 and over, had indeed taken full measurements, but had never fed the data of 3-year-old TV viewers into the analysis. According to Mediapulse, this meant that "around 0.3% of total TV consumption was lost". The error has been corrected since January 30, 2015. Until recently, the internet researcher NET-Metrix, a subsidiary of WEMF and Mediapulse, was also struggling with problems: It was only able to publish the data for January 2015 on March 5 - due to "defects in the hardware responsible for calculating the unique client values", as stated on the NET-Metrix website under the title "Nobody is perfect". A fourth research project that is currently struggling is the poster research SPR+. Although reach data on urban poster sites is still available, the announced expansion to rural poster sites and residents (SPR+ Mob Nat) is already more than a year behind schedule. The reason: checking the huge amounts of data in the single-digit terabyte range is more time-consuming than expected. After postponements to fall 2014 and then to December 2014, SPR+ Managing Director Felix Mende now expects that the validation phase "can be completed during the second quarter of 2015 at the latest". Defects are not always rectified. For example, double counting in the WEMF's circulation certification is deliberately part of the system: the regulations allow publishers to have e-paper subscriptions that are part of a paid print subscription shown as free e-papers, provided the e-papers are used regularly. This means that the same subscription is included twice in the circulation and the latter is therefore inflated (see Werbewoche 19/24.10.2014, p. 47). A survey conducted by Werbewoche among 40 newspaper titles with designated free e-papers shows that some newspapers actively make use of the option to boost their circulation, but there is no transparency: most of them did not provide any information on this.