Measuring technique

Radio consumption in Switzerland is measured rather than being the subject of verbal surveys. The data is based neither on the logs kept by selected individuals nor on the powers of recall of the sampled population. In this way, we can guarantee a high degree of precision. Furthermore, measuring requires absolutely no assistance from the test subjects other than their willingness to take part ("passive approach").

Mediawatch captures the radio measurements by means of a wristwatch with a built-in microphone. This microphone records the sounds of the radio programme that the test subject is listening to. The sounds that have been recorded are converted into digital information in such a way that it will never be possible to listen to it. This digital information is simply a series of numbers which represents a snippet (100 bytes instead of 12,000 bytes) of the original ambient noise. It is saved to a chip in the watch.

The Mediawatch is worn by a selection of individuals representing a cross-section of the Swiss population for seven consecutive days at a time. Following the recording period, the test subjects send their watches back to the partner institute, where their radio consumption can then be calculated by means of audio matching. Every day, over 1,000 measurement devices are in use, which means that radio consumption can be reliably extrapolated to the entire population.

Audio matching is the name of the process in which the data captured by the measuring device is compared with reference data from all the radio programmes being surveyed. This process does not take place in the watch, but in a central computer at the partner institute conducting the fieldwork. At the end of the recording period, all the participants return their watches to the partner institute, where the saved data is retrieved. At the same time, each programme's reference recordings are transmitted to the institute on a daily basis by the stations being measured and a check is made on the completeness of the recordings. As soon as all of the participants' watches have been received for a particular measuring day, the process of comparing the watch data and the station's data – audio matching – gets underway.  This usually happens 18 days after the actual measuring day.

The right of privacy is ensured by the reduction and transformation process applied by the measuring devices when recording ambient sound. The original sound snippets are not recorded as audible sound but merely as number strings indicating changes in pitch and volume. It is absolutely impossible to reconstruct the original sound. This has been corroborated by the Swiss Federal Office of Metrology.

Hardly. Audio Matching works as long as there is an audible signal, even if there is a lot of static. A series of lab tests has ascertained this. Intuitively, this may not make sense, but bear in mind that, unlike the human ear, Audio Matching works on the basis of sound patterns rather than sound. Audio Matching only fails - or, rather, is designed to fail - when the human ear would not be able to make out the radio station either, as in the case of too much ambient sound.

If we allowed sound snippets to match across different time periods, we would produce too many false recognitions. Most likely, a song played on a radio station will also be played at some point in the day on someone's home sound system. If the song is heard at precisely the same moment, we can be fairly sure that the panellist was listening to the song while it was being played on the radio. The Audio Matching allows a lag in the transmission of the signal up to 3 seconds.

In most cases, these are co-produced or shared programmes that we know about in advance. We provide the Audio Matching algorithm with a number of editing rules that decide which of the two stations we assign the listeners to. Generally, they are assigned to the producer of the programme. In the case of co-productions, we assign the listeners to the station they were listening to before the common programme started or continue listening to once it ends.