I was reading about the OJ Simpson book/TV thing that was recently canceled, and the article contained a mention that it was a sweeps stunt designed to bring in ratings for sweeps week. The article went on to explain how advertising rates are based on sweeps, and I stopped reading.
Why, in this day of online Neilson boxes, are advertising rates set based on sweeps ratings? It seems like the whole thing is set up to game the system. Why would it matter how many people watch a particular show during its Ultra Spectacular Episode if they don’t watch it regularly?
I’m assuming that there was some historical reason for doing things this way (like the increased accuracy of having a big survey one time rather than continuous small surveys) But I don’t know enough statistics to know if this is really true. It seems that a rolling small survey could produce more accurate numbers than a single large one, and that any gains from a large one time survey are quickly made irrelevant by the fact that the networks know it’s coming and game the system.
So, are there valid statistical reasons for doing sweeps? Were there ever?
Surveys of viewership are expensive. The more people you include the more expensive they get. Four times a year, the Nielsen people do extra-large special surveys to get at the finer points of viewership. In a regular survey you may only have 12 black females over 65 who live in the Midwest, too small a number to base any extrapolation on. If you take ten times as many surveys that breakdown number becomes large enough to work with, along with the 10,000 other breakdown numbers.
That was the theory. In practice, the networks did indeed figure out that they could game the system by putting special programming into the survey months. This made every number worthless but it was all the numbers they had so everybody played the game.
Sweeps have become far less important in recent years. They haven’t gone away yet but they’re vanishing.
Nobody really knows yet how to measure internet (and cell phone etc.) viewing or what they do to viewership. All is chaos under heaven, and the situation in excellent, as one of Mao’s sayings read.
I wonder how expensive it would be to have those four surveys are random times, or at least not-preannounced times, or at least announced-as-close-as-possible-to-the-survey times. Because the current system is worthless, as you wrote.
The current system seems nonsensical. Everyone knows the numbers don’t mean what they supposedly mean, but they have to have the numbers anyway, so they play along.
A related question: Does Nielsen or whoever does the ratings account now for DVRs and time shifting? I was recruited to be a Nielsen family several years ago and was initially excited to have my viewing habits counted. But when I sat down to fill out the forms, there was no way to do it correctly. It was a pre-DVR (really even pre-VCR) format that required me to note what channel and show I was watching at certain times. But there was no way to indicate that I watched one show on Ch. 5 and taped another and Ch. 2 to watch later, or that I taped Ch. 11 at 10 pm but at 10 pm I was watching something from three days ago.
Try as I might, there was no way to accurately note what shows I watched. If I “watched” two shows that were on at 8 pm Thursday – by watching one in real time and taping the other to watch later – there was no way to indicate that. So the whole thing seemed pointless and I just didn’t do it.
I’d imagine that a big part of the problem is whether advertisers will care what you watched on a DVR. If you don’t watch it live, odds are you’re not watching the commercials.
I have to say that I’m surprised that the ratings system is so slow to adapt. It seems like a large sweep of surveys adds expense (in the form of an unsteady work load requiring temps or overtime, if nothing else) for reduced accuracy. Why didn’t they just change to a constant size pool of surveys covering all times as soon as it became clear that the networks were playing to sweeps? Institutional momentum?