Unlike television audience surveys (which mostly use peoplemeters these days, even in developing countries, radio audiences are still measured with diaries. Some large audience measurement companies, such as one in Switzerland, have been developing wristwatch-like meters for measuring radio audiences, but as far as I know these are not yet being used anywhere in the world for real surveys.
Around the world, the most common type of radio listening diary runs for one week, and is filled in by one person. Usually there's one page opening for each day, with quarter-hour units down the page, and one column for each station. When there aren't very many radio stations in the area being surveyed - less than about 10 stations - the station name is usually printed at the head of each column. To indicate listening to a station during most of one quarter hour, the respondent just ticks box for one station (choosing the appropriate column) on one quarter-hour (choosing the appropriate row on the page).
When there are many radio stations in the area, a common method is to use a system that I understand was first developed in Canada. Each diary comes with a set of stickers, and each sticker has a station name on it. If it's a 7-day diary, there are 7 stickers for each station in the area. When the interviewer is explaining the diary to the respondent, she (the interviewer) asks him (the respondent) which radio stations he normally listens to, finds the stickers for those stations, and sticks them in the blank column headings for each day of the diary. The respondent is asked, if he listens to any other stations during the week, to stick the sticker for that station at the head of a spare column.
In rich countries, the unit of time in a radio diary is usually a quarter-hour. In poor countries, where people are in less of a rush (and often don't have clocks) the time units can be longer. When doing a survey in Ethiopia, I discovered that few households had clocks or watches, and that the times of day had names instead of numbers: there were 13 named times of day, so we produced a diary with a time-name in each row. The problem with using longer time-period is that if people listen to a station for less than half the period, what do they show in the diary? Theoretically, to avoid overstating average audiences, they should leave a space blank unless they listen for more than half the time period. In practice, they don't seem to do that, and this overstates the audiences for stations that people listen to for short periods - e.g. when tuning in only for a news bulletin.
One problem with radio diaries is the definition of "listening". At one extreme, "listening" can mean doing nothing but listen to radio, paying very close attention such as if they are listening to a favorite emcee or language learning program.
At the other extreme, a radio might be audible in the background. Obviously the latter method will produce a much longer duration of average radio listening than the former method. For example, the regular Australian radio diary surveys measured an overall average listening duration of about 3 hours a day, using a definition of "being in the same room as a radio that was audible". But when the government did a time-budget survey, asking people to keep a diary of how they spent all their time during a week, the average time spent listening to radio turned out to be about 10 minutes a week. Naturally, this low figure did not please the commercial radio industry, because the longer the apparent time spent listening, the more able commercial radio is to attract advertising away from television and print media.
When I read the time-budget survey report closely, I found that the definition being used for radio listening (as for everything else in the report) was of primary activity. That is, people spent about 10 minutes a day listening to radio as their primary activity. The rest of the time (the other 2 hours and 50 minutes) they were mainly doing something else, but also had a radio going in the background. In other words, changing the definition made the audience seem 18 times larger. So don't let anybody try to tell you that definitions don't matter!
Thinking of the practicalities of filling in a diary, most people are not going to keep the diary with them constantly and fill it in every quarter-hour. They are more likely to fill it in once or twice a day. That means they will need to remember the times when they listened, and the stations they listened to at those times. This is much easier and more accurate with a loose definition (such as "being in the same room as an audible radio") than with a tight one ("such as doing nothing else but listen to radio"). So for this reason - and also due to pressure from client radio stations to make their audiences seem large - the normal definition of listening is the loose one: being in the same room (or car, etc.) as a radio that the respondent can hear.
Each respondent in the survey represents a certain number of people in the population of the area being surveyed. Because some types of people are more likely than others to fill in listening diaries, and some types also spend more time listening to radio, it's normal for a sample to be divided into a number of small groups - often based on age-group and sex - and audience estimates calculated separately for each group. To arrive at total audience estimates, all the group estimates are added up. For example, women usually co-operate more than men with diary surveys, and old people more than young people. So a typical sample might be divided into 6 groups, three age groups for each sex. Let's say these are 15 to 29 years old, 30 to 44, and 45-plus.
A surveyed area might have a population of 15,000 women aged 45 and over. (This would be known from Census data.) If 150 women aged 45 and over complete diaries for a survey, that's 1% of the population - so each woman in the survey is representing 100 women. If 20 women in the survey listen to a particular station in a particular quarter-hour, they represent 20 x 100 or 2000 women, so the estimated audience for that age/sex group, for that station, at that time, will be 2,000. Repeat this calculation for the other 5 age/sex groups, add the estimates together, and you arrive at an estimate of the total audience for that station, at that time.
Children under about 15 years old usually aren't included in radio surveys. They don't listen to radio much (more time with TV), they have little buying power (so don't interest advertisers), and it can be difficult to get their parents' permission for them to be surveyed.
Of course, because surveys do not cover the whole population, their estimates cannot be assumed to be completely accurate. However if everybody in the population has the same chance of being included in the sample, it is possible to calculate the margin of error. The smaller the sample, the larger the margin of error. For many radio surveys, the margin of error is often around 5%. In other words, if an estimate is 2,000, the true figure is likely to be between 1,900 and 2,100 (because 5% of 2000 is 100).
The data from diaries is converted into several types of audience measure, each of which is useful for a different purpose.The main measures are:
The average audience is simply the average number of people listening to a particular station in a particular time period. This is usually expressed as estimated thousands of listeners, but sometimes as a percentage of the relevant population. (For television, it's the other way round.) The time period can be
These average audience figures can be calculated for any demographic group. In the above example, with three age groups for each sex, there would be average audience figures for each of these 6 groups separately, plus figures for each age group (both sexes combined) and each sex (all age groups combined), as well as a total figure. There would be figures for each main station, with less popular stations combined as "other". For each station and each demographic group, there would be estimates for each quarter-hour of the week, every time zone, each day, and a weekly average. That's a lot of numbers! These tables are computer-generated, and a typical radio diary report may contain 100 or more pages of numbers.
The reach of a radio station, also known as cumulative audience is the number of different people who listen to a station in a time period longer than the basic unit. If the basic unit is a quarter-hour, then (because of the way diaries are filled in, as explained above) the reach for a quarter-hour is the same as the average. As the time period grows, the reach grows too, but m more and more slowly.
Let's continue the example of the audience of 2,000 in one quarter-hour. Suppose the station was broadcasting a half-hour program at the time, and the other quarter-hour of the program had an audience of 2,500. Some listeners would have heard the whole program, while others would have listened only during the first quarter-hour or the second quarter-hour. If there had been a total change of audience halfway through the program, so that nobody listened to both quarter-hours, the reach would be 4,500, because that many different people would have listened to the program. At the other extreme, suppose that all 2,000 people who listened in the first quarter-hour also listened in the second quarter-hour. But because the audience in the second quarter-hour was 2,500, that means another 500 people must have switched on halfway through the program. Therefore, in this case, the reach was 2,500: the number of different people who listened to all or part of the program. So by knowing the two quarter-hour audiences, we can deduce that the reach must be somewhere between 2,500 and 4,500. But to get the actual reach, it needs to be calculated from the listening data for each respondent.
The same principle applies when extending reach from a half-hour, to a long time zone, to a full day, to a week. The longer the period, the larger the reach - but as the length of the time period grows, the reach grows more and more slowly. It can never go backwards - e.g. it's impossible for a 7-day reach to be less than a 5-day reach. Also, because reach must be calculated from the listening records of individuals, the reach figure is limited by the length of the diary. You can't calculate a one-month reach from a one-week diary, for example - because you don't know how many people who didn't listen to a station in that week might have listened in the next few weeks.
Like average audience, reach can be expressed either in estimate thousands of listeners, or as a percentage of the population. It tends to be expressed as thousands for shorter time periods, and as percentages for longer periods - e.g. "FM99 reaches 11,500 people from 7 to 10pm on Saturday night" or "39% of the total population in this area are reached by FM99 at least once a week."
Audience share is a different kind of measure altogether. Both average audience and reach are counts of people. Audience share, though always expressed as a percentage, is not a percentage of people, but of person-hours. It's easy to forget this, but try not to! Take the statement "FM99 has a 40% share of the radio audience in this area." That means: out of every 100 hours that people in the area spend listening to radio, FM99 has 40 of those hours. That does not mean it has a reach of 40%. The reach could be a lot more or (more likely) a lot less, depending on the number of stations in the area, and how long people spend listening to each station.
Think of it like this: if you add together the shares for all stations in an area, the total is always 100%. That's why it's called audience share. It can be calculated for a single quarter-hour, a time zone, a day, a week, or any time period. No matter how few people are listening to radio at a given time (e.g. 4 a.m, when audiences are usually tiny), the share for all stations will always add to 100%. (However, reach figures usually add to a lot more than 100%. If the total of all station's reach figures is 200%, that means the average person listens to 2.00 stations in that time period.)
Share figures are useful for comparing different demographic groups. For example, if a radio station is trying to appeal to young people, it will probably have a much larger share of the under-30 audience than of the age 45-plus audience. Also, because radio audience sizes usually vary in a consistent pattern throughout the day, calculating a station's share for each quarter-hour in the day can show which programs are strongest and weakest, compensating for the variations in total audience size.
Duration is sometimes known as the average time spent listening, which describes it clearly. Consider a time zone audience, which might average 30% of the population. This can be expressed in another way, which is mathematically identical: if 30% of people are listening, on average, then it follows that the average person listens for 30% of the time. So if it was a 3-hour time zone, the average duration is 30% of 3 hours, which is 90% of an hour, or 54 minutes. I mentioned earlier that the average time spent listening to radio in Australia is about 3 hours a day. That's a duration figure. It can be converted back to an average as follows: since there are 24 hours in a day, and the average person listens for 3 hours, the average audience is 3 out of 24, which is 1 out of 8, which is 12.5%. Therefore the average radio audience in Australia is 12.5% of the population. You can easily switch between these two figures, depending on which is more useful for the point you are making.
That's the simple definition of duration, but often it's more complicated that, because sometimes you want to exclude people who didn't listen at all. That is, the above definition calculates an average of the whole population, but sometimes you want an average duration for those who actually listened. In that case, the average audience must be divided by the reach, then multiplied by the length of the time period. For example, about a fifth of the people in Australia don't listen to radio at all on an average day.That means the all-station all-day reach is 80%, not 100%. if the average is 12.5% of all people, it must be 12.5/.8 of the actual listeners, or 15.625% of them. Therefore the duration (among those who actually listen) is 15.625% of 24 hours, or 3.75 hours.
Impressions, also known as impacts is a measure used by advertisers. It's the sum of the audiences at specified times - e.g. when ads are broadcast. If a station has an audience of 2,000 in one quarter-hour and 2,500 in the next, and the same ad is broadcast once in each quarter hour, that will be 4,500 impressions: the number of times the ad was heard, regardless of the number of different people who heard it. If the ad is broadcast twice in each quarter-hour instead of once, the number of impressions will be 9,000. If the total cost of advertising is divided by the total number of impressions, you can calculate the cost per thousand impressions. This is a measure that advertisers find useful when working out which stations to advertise with.
Frequency is another measure used mainly in advertising. it's an answer to the question "how often did listeners hear the ad?" If (continuing the previous example) an ad is broadcast twice in each of two quarter-hours, the frequency will range between 0 (because some people won't have heard it at all) and 4 (among those who listened to the station for both quarter-hours). Actually, there are two measures of frequency: the average frequency and the frequency distribution. Frequency is normally based on the whole population - because that's who the advertiser is trying to attract. If we know that the reach for the two quarter-hours was 3,000, and that there were 9,000 impressions (as above) the average frequency must be 3.
A full frequency distribution would show how many people heard the ad 4 times, 3 times, twice, once, and not at all. These figures cannot be calculated from the data I've supplied - you'd need to go back to the raw data. A common goal for advertisers is to achieve what they call a 3-plus reach - that is, the number of people who hear a message at least three times. The reasoning is that people don't fully understand a message till they've heard it 3 times - which is one consequence of defining radio listening as being within earshot of an audible radio. If you used a tighter definition, you'd find that one hearing was often enough to be effective. The 3-plus rule of thumb is only a starting point. It is not well supported by research data.
If all listeners to a station listen only to that station, its loyalty is 100%. This is usually measured over one week: a standard cycle time in radio listening, both for programming and for survey samples. By definition, loyalty cannot be zero: listeners to a station must spend some time listening to it, otherwise they wouldn't qualify as listeners. So the formula for loyalty to station A is:
(time spent listening to station A) divided (by time spent listening to all stations)
...usually expressed as a percentage. In an area with between about 5 and 20 stations available, a loyalty figure of 50% is on the high side. In other words, if the people who listen to your station at least once a week spend more time with your station than with all others combined, you are doing well. Unfortunately, because of the principle of double jeopardy that applies so widely to most consumption behaviour, the smaller the audience share your station gets, the smaller your loyalty is also likely to be. Statements such as "though we get a tiny audience, they are very loyal" are - unfortunately - hardly ever true. Occasionally, loyalty is called "share of ear".
Those are the main measures of radio audiences, but others are sometimes used. One example is unduplicated reach - the number of people who heard a program or ad only once, even though it was broadcast several times.
Some audience measures can be done without a diary - e.g. the survey question "Which radio stations have you listened to in the last 7 days?" will usually produce a set of answers very similar to the weekly reach figures from a 7-day diary. However, audience share is much more difficult to estimate by simply asking people - but diary surveys are very expensive, and require complex software. By testing various question wordings, and comparing them with diary surveys, I found a question that closely approximates share figures without doing a diary: "Which one radio station do you listen to most often?" As with the share figures, the answers to this for all stations add to 100% - after you've ignored the people who can't answer and those who insist on giving two answers. Results are usually within 2% of share figures from a diary survey, but slightly favour stations that people don't spend much time listening to.
Though radio diary surveys are done regularly in most western countries, the results are not often published. Because surveys are expensive to do, and the results are valuable for radio stations, the survey reports are sold privately for high prices, not published. Therefore, little information on audiences is available on the Web. However, findings are so specific - particular stations, in particular places, at particular times - that they're not usually of much interest outside their local context. See our Radio research findings page for some general conclusions, that we've found to apply in many different situations.