How do OzTAM Ratings work?
OzTAM boss unravels a little of the way people meters work in the home.
- Published by David Knox
- on
- Filed under News
If you’re interested in how the OzTAM ratings system works, CEO Doug Peiffer chats with News Corp about some of the intricacies and myths, today.
Australia has 3500 people meters across Sydney, Melbourne, Brisbane, Adelaide and Perth, and another 2500 in regional areas (Regional TAM).
“Our households are asked not to disclose their identity. It means their tastes can’t be targeted and played to,” he explains.
“The 3500 households in the metropolitan market translate to more than 10,000 eyeballs, and with another 2500 households regionally, it’s one of the largest samples — in terms of panel size relative to overall population — in the world.”
You can read more about how you cannot request to join the panel survey, and how users have to log in when viewing at home, to acknowledge just who is viewing.
7 Responses
I seem to recall one night when ACA had absolutely skyrocketing rating when they promod the show by saying ‘If you have a people meter you can’t afford to miss tomorrow night’s ACA’. Rules changed shortly thereafter
Have often wondered how to lure people with meters to TVT! It would revolutionise the site. Open to any suggestions!
My suggestion: Don’t; at least not without talking to OzTAM first. It may seem like a little thing, but could quite easily have an effect large enough to show up in their measurements.
Even adding 2,500 in the country
3,500 reflecting the taste of 25,000,000. To me it doesn’t compute and never has.
There are many things I’m ignorant of – doesn’t mean they’re not true or accurate or whatever…
But statistics isn’t one of them, so here goes :-
Lesson 1: Know the difference between what you’re sampling, what you’re measuring, what you’re trying to measure, and how to get from A to B to C. For example, OzTAM samples households not people – so, for a start it’s (3500+2500) from ~9 million, not ~25 million.
And by about Lesson 6 you’d know that such a sample is likely to be about 99% accurate – it gets to a point where adding more people to the sample doesn’t make it any more effective, so it’s not worth spending money doing so.
The accuracy is such that for a show with a reported rating of 1,000,000 that the actual rating will fall somewhere between 990,000 and 1,010,000 – so basically within +/- 10,000, which is close enough.