How are TV viewing statistics obtained?
My quest to gauge the length of time ColdFusion developers have been developing in ColdFusion has got me exploring all avenues as to how to obtain figures that actually mean something – it seems that I don’t actually have to!
Take for example those adverts on TV for various female products for eyelashes, lipsticks and shampoos – it seems that it’s enough to sample 100 women to claim that their product is the nations favourite!
TV viewing stats are another area that intrigued me – claims that 7 million people watched last nights EastEnders- where does that figure come from?
It turns out that in the UK BARB (Broadcasters audience research board) are behind these figures. Out of the 25+ millions homes here in the UK, 5100 homes make up the reporting panel – a mere 0.02% of UK homes!!!! Monitoring is automatic by a device connected to TVs, PVRs, VCRs etc which registers what is being watch and household members use a ‘peoplemeter’ to register when they are in a particular room – on a daily basis around 11300 make up the sample size. So what happens next, well there’s no specifics but I imagine the sample set is extrapolated up to the known population of the UK (just under 61 million in July 2008) – so therefore a mere 0.018% of the population produce the viewing figures for the entire nation. Viewing statistics are then available after a nightly data retrieval from the device and made available on the BARB website.
I had no idea that the sample set was so small – used to produce a figure so highly regarded but I guess that’s the whole thinking behind ‘statistics’ and random selection of parties to obtain a true picture.
The quest continues….
No related posts.
Similar to how Nielsen does it here in the US, but Nielsen uses a much larger sample size:
http://en.wikipedia.org/wiki/Nielsen_Ratings
My father worked for Nielsen Media Research for awhile as an engineer targeted on automated collection of tv viewing habits (they also measure radio, billboards, and with a patent credit to my pop, can even track magazine ads all without the user’s active participation).
Even more fun, is that there are version of the peoplemeter that know if they are not attached to a person, and I believe that they can track even if you’ve fallen asleep.
I reckon that the downside to these, in many cases, pretty cool engineering accomplishments is that they lay the groundwork for a lot of intrusive metering for things beyond just television viewership.
Radio statistics are interesting too. For one month, they have a number of people complete a paper survey of all the radio stations they listened to, when and for how long. I don’t know how many people they use. I got asked to do this once years ago.
The month they use is May. Remember Radio 1’s “31 Days in May” competition bonanza they always used to do…. now you know why!