Michael Baxter: Television Monitoring
24 January 2019

Television monitoring, meaning measuring the number of people watching TV in any given minute or during the course of a programme, is conducted in most countries.  It is important for TV companies to know which programmes are popular so that similar ones can be made and broadcast.  It is also important for advertisers to know how many people, analysed by age, sex and other characteristics, have watched any given advert or are likely to watch it next time it is aired.

Regular monitoring has been conducted in many countries for several decades, but the technology is rapidly improving.  The original method was to ask people to fill in paper diaries of what they watched, but modern electronic methods are better and impose far less response burden.  Every country is unique, and local knowledge is important in ensuring that monitoring is as good as possible.  In a few countries, there may be political issues.

Audiences are getting more fragmented.  As recently as the early 1990s, there were only four TV channels in the United Kingdom.  Now there are hundreds, although many have such low viewing figures that only rough estimates of annual viewing can be made.  Also, as well as TVs, people are increasingly watching on PCs, tablets and mobile phones.

The usual procedure for setting up a monitoring service is for the TV companies and advertising agencies to set up a Joint Industry Committee (JIC) and issue invitations for companies to bid for the work.  In the UK, the JIC is called BARB (the British Audience Research Bureau).  However, in a few countries, notably the USA, a market research firm collects the data on its own initiative and sells it to interested parties.

Monitoring is conducted using a panel of private households.  (Viewing outside private households, for example in communal establishments or pubs, is generally ignored.)  Obviously, the larger the panel, the more reliable the results and the greater the disaggregation of data that can meaningfully be produced.  However, a large panel is expensive.  In the UK, the panel size is 5000 households.  However, in most countries the panel is smaller.

The first step in setting up the panel is to conduct an establishment survey (ES), a large household survey to gather information on characteristics of the population such as how many TVs they have and how many hours of viewing they have per week.  As with all household surveys, there is a risk of non-response bias.  To reduce this, Kantar Media grosses up the results to the estimated household population by age and sex, and the estimated distribution of household sizes.  It uses a modified RIM methodology which I have devised.

Panel households are then recruited from among those households which responded to the ES, since these households have known characteristics.  To ensure that the panel is reasonably representative of the population, a CHAID analysis of the ES using claimed hours of viewing as a dependent variable.  This shows which characteristics of households are likely to affect the amount of viewing, and the panel is selected so that the distribution of such characteristics in the panel is as close as possible to the distribution among all households.  For example, with a panel of 1000, if it is found that 75% of households have one TV, 20% have two and 5% have three or more, then the panel should aim to have 750, 200 and 50 households respectively with that number.  Obviously, households with no TVs are excluded.

The panel must be continually updated.  Households may wish to stop being on the panel or may be discarded if they produce bad data.  Characteristics of households may change over time as people are born, or leave home, or die, or just get older.  Further, the characteristics of the population change over time; in particular, the proportion of households with different reception types (e.g. terrestrial, satellite, cable) may change rapidly.  Regular ESs are needed to monitor this and to provide a new pool of households for recruitment.

For conventional TVs, a physical monitor is attached to each TV.  For PCs, etc., software (virtual meters) may be installed on each item.  This means that out of home viewing will also be recorded, and there is a risk that new items will not get the software.  A better procedure is to install a meter on the router.  In some countries, data are available on the number of views of each programme on the TV company websites, such as the BBC iPlayer.  However, this gives no information on the demographics of viewers, so this must be combined with the panel data to get the best results.

Both viewing methods and monitoring technology are likely to continue to evolve in the coming years.

Michael Baxter is Director of Statistics for Kantar Media.  His predominant area of responsibility is television audience measurement, across dozens of countries worldwide.  He is also involved with the Target Group Index.  As a former Government statistician who was at one time responsible for producing national population estimates and has been Chairman of the Official Statistics Section of the Royal Statistical Society, He can discuss official statistics from the viewpoint of producers as well as users. Michael has been a member of the MRS Census and Geodemographic Group since January 2017.