Audience Dialogue

Measuring radio and TV audiences with telephone surveys

Case study 9

In the past, radio and TV audiences were measured mostly using diary surveys. Randomly selected households received diaries, and were asked to fill these in for a certain time - usually a week. Sometimes the diaries had a few extra days at the beginning, for practice. These initial days were not counted. Sometimes, too, the diaries had an extra day at the end which was not included in the survey - because a lot of respondents seemed to lose interest in recording their viewing or listening on the last day. So the diary may have looked like a 10-day diary to respondents, but only 7 days (the 3rd to 9th) were actually used as real data.

For TV surveys, there was usually one diary per household - because (in those days in the distant past, at least 10 years ago, most households had only one TV set. (We are talking here about the richer countries: North America, Western Europe, Australasia, and Japan.)

For radio surveys, there was one diary for each person. Children aged under about 10 were excluded, for several reasons: they don't read well, they don't listen to radio much, and they aren't of much interest to advertisers.

Around 1990, TV meters gradually took over from diaries. With a TV meter system, a panel of households is used, selected at random, to be a good cross-section of the population. Within the selected households (which are usually given minor incentives to compensate for their trouble) a technician visits, and connects each TV set in the household to a meter, and all the meters to a phone line. The meters record which channels each TV set is tuned to, at which times. In the early hours of the morning, the master meter in each household automatically dials the research company's computer, and downloads the viewing information for the previous day. Next morning, when the researchers arrive at work, the computer has calculated the previous day's TV ratings. A table showed the percentage of households which viewed each channel, in each quarter hour.

Later, "people meters" were developed. These meters sat on top of each TV set in the panel households, and were wired to the TV sets. In the front of each meter was a set of buttons (one per person) and lights. When somebody switched the TV set on, all the lights would flash. Of course, this was very annoying, and the only way to stop it was to press one or more buttons on the meter, to show who was watching. With this sytem, each person in the household has their own button (with a light inside it). The meter is programmed to know the age group and sex of the person who owns each button. For example, button number 1 might be recorded as a woman in the 45-49 age group. When that button was pressed, a person of that sex and age group was recorded as watching. The connection between the meter and the TV set recorded the channel tuned in at the time.

Every 45 minutes, if nobody had pressed a button to indicate that they had started or stopped watching, all the lights on the meter would flash, and everybody would have to press their button again, to show that they were still there.

This method works quite well - as determined by occasional telephone surveys, which are used to check the meter results. The main problem is that people forget to unpress their buttons when they leave the room for a few minutes. But the latest generation of peoplemeters solves this problem: instead of buttons, they have a video camera, which has been taught the shapes of the faces of everybody in the household, and can recognize them automatically. I don't think these are yet in wide use: they are very expensive,and hundreds of them are needed.

In the middle of the night, when the modem built into the peoplemeter contacts the central computer, it sends a message such as "woman aged 45-49 watched channel 2 from 7.43 to 9.11pm".

With peoplemeters, we research executives received much more detailed printouts. Instead of just seeing that 37% of households were watching channel 2 between 7.45 and 7.59pm, we'd also find out how many people were watching, and which age groups they were in. This information was very useful for advertisers, who often wanted to reach a particular "target group" with their commercials, and sometimes of interest to programmers.

Meters work well enough with TV sets, which are large, and don't get moved around much. But for radios, which are a lot smaller, and don't needed to be connected to the electricity supply or large aerials, I don't think meters have been developed yet - though people have been trying for years. When I was working for the Australian Broadcasting Corporation, the managing director was very impressed with TV meters, and ordered the engineering department to produce some radio meters. They worked on the project for several years, but in the end had to give up. The best prototype they produced was a meter about the size of a packet of cigarettes, which would be sticky-taped to a radio. It would detect what frequency the radio was tuned to, and store this in its memory, along with the exact times the radio was switched on and off.

But what if the volume was turned right down? With some clock-radios, the tuner is always on, and such a meter would record them as being listened to 24 hours a day. So the meter needed to have a built-in microphone, to detect if there was any volume. The problem was that we couldn't find a small enough microphone. The smallest was one used in medical research: it was designed to be put down people's throats, so that doctors could hear their stomachs rumbling in stereo.

The idea was that interviewers would deliver these small meters to randomly selected households, sticky-tape one meter to each radio, and come back a week later to retrieve the meter. It would then be inserted into a special machine to read its memory, showing which stations it was tuned to during that week (and had the volume turned up to an audible level).

After a few years, this project was cancelled. It was costing a lot of money, and hadn't produced any usable meters. Later I discovered that radio meters had existed, many years earlier. In the book Successful Television and Radio Advertising,by Seehafer and Laemmar, publsiehd in the USA in 1959, there's a photo of a radio meter, dating from the late 1940s. It was about the size of a loaf of bread - but in those days, radios weren't portable, and were usually about a metre high.

So much for radio meters. The industry is still working on them: the latest attempt I heard of looked like a wrist watch. People in the sample would wear this, and it would record all radio signals they heard during a day. (Radios re-transmit high frequency waves when switched on, so the device wouldn't need to distinguish between a real-life conversation and a talkback program.) Each night, the panel members would put their "watch" in a special device connected to a telephone line. It would transmit their daily listening details back to the central computer, just as a TV meter does. I don't think these meters are in use yet, but probably within a few years, they will be.

So, for TV, diaries gave way to meters. For radio, diaries are still the most commonly used method of recording audiences. In some countries, people had used telephone surveys to measure radio audiences, but these usually produced much smaller audience estimates than the meter surveys. Typically, a radio diary would produce an average listening of about 3 hours a day. A telephone survey, measuring the same thing, would produce only about 2 hours. Naturally, the advertisers did not like this.

In Australia, though I was working for a non-commercial radio network, we bought the diary surveys carried out in the large cities by the AGB McNair Anderson group. This was much cheaper than doing our own surveys, and we were reasonably satisfied with the quality of these diary surveys.

However, the AGB group did not carry out radio surveys in country areas. These surveys used the weekly diary method, which requires interviewers to visit households to leave diaries, and again to collect the diaries. The population of rural Australia is thinly scattered, and travel costs for door- to door surveys would have been very high.

The regional radio network I was working for placed great importance on its country listeners, and wanted to know the audience sizes. Nearly all country households had a telephone by then (late 1980s) so we considered using telephone surveys. I thought that the reason for the low audiences recorded in telephone surveys was that people often listen to radio as a background activity, and couldn't easily remember their radio listening, even on the previous day.

So we developed a "yesterday and today" diary, in which respondents were first asked to remember back to "the time you woke up yesterday. Did you listen to radio soon after you woke up? What time was that? Which station? When did you stop listening to that station? What did you do next?"

By asking questions like this, and finding out where people were during each part of the preceding and current day, we developed a system which averaged a little over 3 hours a day radio listening: slightly more than the diary surveys. We found that we couldn't use a normal questionnaire: there were too many possible combinations of answers. Instead, we gave our interviewers detailed training in probing respondent's activities and locations. If they didn't know which station they were listening to (as was often the case) the interviewer would find out exactly where they lived, and prompt them with the call signs and frequencies of stations broadcasting in their area.

Though we found the same total listening as the diary surveys, our telephone surveys found more listening sessions. We suspected that with diaries, people wrote down what they thought they normally did. With our surveys, the interviewers were trained to ask what they actually did, "today" or "yesterday". Though we can't prove it, we think that our method produced more accurate data than diaries - because the interviewers engaged in a dialogue with the respondents, questioning and correcting their answers as they were given.

The big disadvantage of the "yesterday and today" diary method is that many people who listen to a station only occasionally (e.g. once a week) are not included in the recorded audience. A written diary, if it runs for at least a week, can produce a "weekly reach" audience figure: the percentage of the population who listen to the station at least once a week. A two-day telephone '"diary" cannot do that. Instead we had to ask directly "Which stations have you listened to in the last 7 days?" - which produces slightly inflated answers. Due to the telescoping effect of memory, what some people remember as being 7 days earlier is often more than that.

In countries where few people have telephones, and the literacy rate is not high, our method could easily be used in face-to-face surveys. We have done this in Vietnam, with good results. The only drawback with this method is that interviewers must be trained much more thoroughly than with normal diary methods - in which the interviewer simply leaves some diaries to be filled in, and collects them a week later.

- Dennis List