Some of the variables used in the current weighting matrix in SAARF’s Television Audience Measurement Survey will change today [6 August]. SAARF has announced that the interlaced RIMs will be unbundled in a bid to increase data stability and give users more freedom with the data.
The TAMS sample data is weighted back to the total television viewing universe using the reiterative method (RIM) of weighting. This method places household and individual variables into so-called RIMs, which are weighted to represent the total television universe.
“Our test data using non-interlaced RIMs shows that by ungrouping the interlaced RIMS, the range between the minimum and maximum weights has been reduced,” says Dr Michelle Boehme, SAARF’s technical manager. “This means that the difference between the minimum and maximum number of people in the universe that one person in the sample represents has decreased. There is also more stability within the main target markets of LSM, age and language.
“We believe the added stability that comes from unbundling the variables, and the fact that users can now combine the weighted variables any way they see fit, make these changes to the TAMS RIM weighting extremely worthwhile.”
Boehme adds that the size of the SAARF TAMS panel will soon start to increase, in the run-up to the new seven-year TAMS agreement. “The number of reporting households will virtually double, increasing the total number of individuals to more than 10 000, which could make other variables large enough to weight separately too,” she says.
A number of these RIMs are interlaced, meaning they have more than one variable included in them. For instance, TAMS currently groups the language demographic and the gender demographic.
The variables in the TAMS RIMs, both household and individual, have been tested to ensure they are the ones that have the greatest influence on viewing behaviour.
These weighted demographics, many of them interlaced with each other, are LSMs, age, gender, language, province, community size, access to satellite TV (and other Pay-TV) or a PVR device, and whether the viewer is “homebound” (such as retirees or housewives) or “out and about” (all other work statuses).
When segmenting markets using these weighted variables, users can know they are getting an accurate, representative view of the total market.
SAARF discovered however, that the interlaced RIMs opened users up to inadvertently working with unweighted data, when they created markets which did not match the predefined RIMs.
For example, the LSMs are currently grouped as LSM 1-5, LSM 6, LSM 7-8, LSM 9, and LSM 10. If a user’s target market was LSM 6-8, he or she could group LSM 6 and the grouped LSM 7-8. The results would still be weighted, since both pre-created groups exist on TAMS.
If however, that user’s target market was LSM 6-7, to group LSM 6 with LSM 7 – a variable which is weighted only in conjunction with LSM 8 – the results would be unweighted.
SAARF conducted various tests to determine how best to solve this problem, and has concluded that the current interlaced RIMs should be split out. All other variables that were used separately before in the RIM matrix will remain as is.
This will bring more stability to the TAMS currency, while giving users more freedom to work with the data without being bound to SAARF’s predefined groupings.
The changing individual RIMS
The LSMs will be unbundled from each other and placed into separate RIMs, with the exception of LSM 1-4 which must remain packaged together due to the small sample sizes of the individual LSM groups at this end of the market.
The LSMs will also be separated out from the weight of viewing (“homebound” or “out and about”).
This means that LSMs 5, 6, 7, 8, 9 and 10 will now be weighted separately, ensuring that any combination of them will be weighted as well.
Language was interlaced with two demographics: gender and age (grouped as per the industry norm of 4-6 years, 7-10, 11-14, 15-24 and so on).
These RIMs will all be unbundled, and the languages which are big enough – Afrikaans both, English other, Nguni and Sotho – will now be weighted separately.