TV Audience Measurement

Every year, Finnpanel conducts an Establishment survey where we study TV device configurations, signal reception and known channel penetration in Finnish households.

Since 1995, the Establishment survey has been conducted as a postal questionnaire. The survey is conducted twice a year using a random sample of 12000 households from the Population Information System.

One of the functions of the Establishment survey is to provide a sample source for the TAM survey. The sample of the Establishment survey must be good enough to provide public statistics as well as control data for the TAM survey. The Establishment survey describe the population regarding factors that are essential to TV viewing. These factors are for example type of signal reception, ownership of equipment used for time-shifted viewing, type of tv-set and channel (pay and non-pay) penetration.

The sample of the TAM survey consists of 1000 households totaling approximately 1950 respondents older than 3 years. These respondents are weighted to present the 3+ population in Finland. One respondent represents approximately 2800 other people of same age and living location.

The representativeness of the panel is monitored constantly.  The structure of the sample is continously compared to demographic information. This is an important issue that affects the reliability of a panel survey; the panel sample can be fixed if it becomes distorted.

Because the TV audience measurement is based on a random sample not anyone can join the panel. The Establishment survey questionnaire is mailed twice a year to 12 000 people randomly selected from the Population Information System. In the questionnaire the respondents are asked if they are interested in joining the panel. If they are willing, their data is included in a database out of which we select new households when needed.

The information we get from individuals in the panel are handled with care. Every panelist has got her/his own identification-ID based on which the data is analysed. We do not disclose personally identifiable informations about panelists to anyone outside Finnpanel.

Both heavy and selective TV consumers are presented in the panel. The amount of viewing correlates with age — older people spend more time watching TV than the younger ones. The Finns watch TV approximately 3,5 hours a day. On an average day, approximately 15 % of the panel members do not watch television at all. 

A meter is connected to every TV-set in the household. There is also a remote handset that panelists use to declare their presence when wathing TV.

Online video content is measured with a meter which is connected to the household's router. 

When a meter is installed to the household, Finnpanel's technician provides guidance for using the meter and gives the household a user manual for the meter. After some training the household is included in the panel.

When the TV set is switched on, the meter display prompts "Who Is Watching?". Household members that are present declare themselves by pressing their individual presence button in the remote handset. A presence indicator for each person is shown on the meter display. When they leave the room they press the button again and the presence indicator clears. Each household member has her/his own button on the remote. The guests of the household also declare their presence using the remote when watching TV.

Pressing the presence button is all that needs to be done.

The panel members are instructed to use their presence button whenever they are present in the room and able to watch TV.

The households receive an incentive of 180 eur per year if there is only one TV-set in the household. The incentive is higher for households with more than one meter. This income is subject to taxes.

The yearly panel turnover rate is 20-25 %. In practice, this means that 20–25 households are replaced each month. It is important that the panel structure follows changes in the population and TV issues. For example, when the number of one-person households has increased and accordingly the  number of larger housholds has decreased. When the amount of households with several TV sets has been increasing the amount also changed in the panel.

The measurement is accurate to the second and includes data about the channel and which members of the household are watching. The meter clock is synchronized each day during data collection.

Continuous viewing of a channel for at least 15 secods is registered. The second by second data is consolidated to minute data. The channel watched during the last second of the minute is credited to the whole minute. The viewer estimate for the minute is calulated by summing all viewers in that minute.

The transmission logs from the TV stations are transferred to Finnpanel during the night. The logs contain names for all programs and spots and the exact start and end times. This data is matched by time to the viewing data from the meters.

The panel households are required an over-all commitment to the survey, confirmed by a signed agreement. All panel households are guided personaly in their own homes during the meter installation. Participation does not mean that viewing habits should be changed in any way. 

The panel households also receive a user manual for the meter. Despite all this, it is clear that sometimes someone might forget to press their presence buttons. In order to minimize the errors, the meter constantly displays the ID letters of all members (and guests) that have declared their presence.

Finnpanel regularly contacts the panel housholds and remind them of the importance of using the presence buttons.

The system has built-in checks discarding faulty data. For instance:

1. Events where the TV set had been switched on without any viewers declared are automatically discarded. 

2. Events for unknown household members are discarded.

3. One person can watch only one TV set at a time. If a person declares herself/himself to several sets, only that latest declaration is acknowledged.

4. Panel members that have extensively long viewing statements or do not watch at all during a certain period are listed and the reason for the anomaly is investigated.

If these anomalies are received continuously from the same household the household is contacted to find out the reason. It can be a techical failure in the meter but sometimes it is due to ignorance by a panel member. If the ignorance continues, the household is removed from the panel.

There are two main methods for reporting TV viewing

1. An overnight report is produced each day. The report contains ratings for previous day's programs of all major TV channels. The demographic group of the report is 3+ years old population including guest viewing. The ratings and shares are calculated for each program as well as for 15 minute periods.

2. Additionally there are software tools to analyze the data to produce ranking lists, campaign analysis, frequency analysis etc.

The fllowing figures are calculated:

Estimate

Number of viewers (in thousands).

Cover %

Number viewers proportional to the population of the target group.

Profile %

Proportion of the target group's viewing of the programs total audience.

Share %

Channel's share of total TV viewing.

Reached Audience

Reached audience of 3+ population, either in thousands or %.

Average Viewing Time

Time (per person) spent viewing during a day

The rating for the program is the average of minute ratings during the program.

Example: Program (10 min)

Minutes
 1        220 000         
 2        220 000         
 3        225 000         
 4        240 000         
 5        250 000         
 6        250 000         
 7        255 000         
 8        270 000         
 9        280 000         
10        280 000         
Sum     2 490 000

Rating for the program: 2 490 000 / 10 = 249 000

As in all surveys based on sample the figures reported are not quite exact. The size of the sample has a direct effect on the margin of error. In the tv audience measurement  the margin of error is approx +- 2 % units. So if the audience of a program is 300 000 , the margin of error is about +- 50 000.