TAMS News Releases

Issues Which Have and May Affect TAMS Ratings during 2006

As quite a number of improvements were and are still being made to the TAMS Panel, the SAARF TAMS Council requested SAARF to inform users of the changes. Also of problems that we experienced which are out of SAARF and AGB Nielsen Media Research’s control.

In debating the improvements and upgrades to the TAM panel, it became evident that the communications industry requires an overview of how ratings may have been affected during 1st quarter of 2006. Furthermore, there are other improvements planned for later in the year which will also have a ratings effect.

TAM panels are universally used around the world to estimate TV ratings. The technology is proven and well documented. Our current contractor AGBNMR measures TV audiences in 30 countries around the world, including the UK, Australia and Italy. However, the estimates are sensitive to change, which is understandable.

In this document, we are listing the various past and future events that could have or will have an effect on the ratings, both decreases and increases.

1. Change of Processing Software to Pollux in Week 01 of 2006

Central to every TAM operation is the processing software. This software accumulates and processes the viewing data from panel households. This is a complex procedure and happens every single day. Firstly, households are validated for example whether the household could be contacted (polled) or not; was the meter working correctly; was there too much viewing with no one logged in; etc. These rules are called the “Edit rules” and are used to validate each household everyday to determine which household can be regarded as ‘good’ to report on.

Once the good households are validated, their data are weighted up to reflect the TAMS universe. At this point any skews in the sample are effectively re-weighted back to proper representation.

The order in which weights are applied, the exact number of rejected or good households and the application of the edit rules, all contribute to the ratings calculation. Simply put, 1 piece of processing software cannot possibly duplicate another.

The SAARF Tender Committee approved the change of processing software, because Pollux is “state of the art” software for Overnight or Day-After ratings, as it has numerous quality control checks to ensure data integrity and also panel management control.

From August 2005 to the changeover to Pollux in January 2006, we went through a protracted period of testing Pollux output against the previous software. Yes, there were differences in the ratings calculation, but over time these were minimised. There is no doubt that the changeover to Pollux at the beginning of 2006 constitutes a small rating change.

2. Power Failures in the Western Cape

After the severe failures during the last 2 weeks of February, we analysed the viewing data of the Western Cape, and it was very clear that this period, compared to previous weeks, showed much lower ratings. The effect was significant.

We need to add that those households on the panel that cannot watch TV because of an failure, must stay on the panel, so replicating the effects of the failure on viewing levels in that area. Clearly, we cannot remove these households and so boost ratings as it would be indefensible as viewing did not take place in those households, reflecting reality.

Currently, reactor 1 at Koeberg is coming back on line, and Reactor 2 is about to be shut down because re-fuelling must take place. This is a lengthy process. Cape Town needs both reactors running in order to produce enough electricity during the cold weather. Depending on the re-fuelling process we could see severe failures again - and these will depress ratings again.

The real problem may well be in Gauteng. In previous years during failures when whole sections of Johannesburg went down, we did not specifically track the effects. If, as forecast, we have serious failure in Gauteng this winter, there would be a significant depressing effect on ratings, remembering that Gauteng has the bulk of our panel, in line with population densities.

3. MTN Problems with GSM Polling

We use GSM enabled polling in many households and not Telkom landlines. This is biased to upmarket DStv households who will not allow us to use their landline facilities. In the new rural panel we have no choice but to use GSM because Telkom availability in rural households is so poor.

From 08 April to the 20 April 2006, MTN’s GSM polling effectively stopped working. MTN experienced a load problem at its base station in the Ormonde area (which AGBNMR use), and decided to compress voice mail in order to accommodate the extra load. Unknowingly, they disabled the data transmission facility through the compression. It took MTN until the night of 20 April to find and rectify the problem.

This affected hundreds of panel households. However, we managed to re-route a substantial number of households through Vodacom, where it was possible. Data was also accumulated in the affected households, which we could poll once MTN was up and running again. However, we have no doubt that ratings were affected over this period. Most probably negatively, because those households with no polling would have been included on the reporting base for a week – the normal time period allowed to establish reasons for households being down, and to rectify the problem.

4. Easter Holidays 2006

April this year had the Easter weekend, and then directly after that another very long weekend again. Every TAM panel around the world measures TV viewership in home. In other words when households go on holiday there is no viewership, because the members are not at home. Households on holiday must be left on the reporting panel.

Including households with no viewership depresses ratings and gives a measure of who is on holiday.

5. Rural Panel

As per the new TAMS contract, AGBNMR is currently installing the rural panel, that has not been measured up to now. This will not have affected ratings as yet, because the rural households are not in the reporting base. However, once we have reached “critical mass” in terms of reporters, we will begin to test the effects of weighting these households up to universe.

This will entail re-weighting the whole TAM panel and again, this will affect ratings. Obviously, every care will be taken, together with the TAMS Council to minimise the affects, but there will be a ratings effect, if only minor.

6. Measuring all TV Sets in House and DualView Decoders

There are 2 upgrades that will positively affect ratings, and probably significantly.

Previously, it was virtually impossible to measure every TV set in house, because the various master and slave meters were connected via cabling. Households usually refuse to have more cabling tacked to their walls, and in some cases in concrete roofed flats / apartments or houses, the additional cabling was actually impossible to install.

As part of the new contract, all the TV meters now in use, must be connect via radio links. This means that through time we will eventually measure virtually all TV sets in panel homes. The effect of this will increase the number of channels viewed, maybe also the time spent viewing TV in those households, and this will positively affect ratings.

We have fairly recently found a fool proof method of measuring both output channels of the DualView decoder. This, together with the full measurement of every TV set in panel households, again means an increase in the time spent viewing TV and an increase in ratings.

Both these effects will increasingly be felt as more DualViews come on the panel, and more additional TV sets are measured. Both these upgrades are currently underway.

We hope that this communication gives more information on how upgrades to the panel, as well as methodological changes can affect ratings. The worrying issues are things which we cannot control, such as the power situation in the country and polling problems due to network communications failures.