In the online world, 1,000 impressions means that an ad got downloaded (or at least the request for a download came in) 1,000 times. This isn’t the case in traditional land.
If a TV show gets a viewer, even for a minute, it can count as though the viewer watched the entire program. That means that people hanging around a show just long enough to determine how terrible it is can be charged against unwary advertisers.
Television audiences are calculated by measuring a selected group of people and extrapolating what the rest of the world is likely watching. If I’m a channel surfer who’s being watched by the Nielsen people, I could be counted as having watched eight different two-hour programs all in the same two hours. This time warping effect comes about because the Nielsen People Meters measure each minute of use. Most figures I’ve seen count the middle minute of each quarter hour as the determining channel. That methodology allows for a four-fold multiplication of viewing reality.
At its theoretical extreme, someone paying a $4 CPM could actually be buying a $16 CPM, although I’m sure the effect is nowhere near that great.
In most cases on the Internet, an ad gets downloaded to a users page immediately following the user having clicked on something. That means that there’s usually a human on the other side. In the TV world, there are a lot of people visiting the kitchen, bathroom, vacuuming, or just plain leaving the TV on. A study a couple of years ago showed that a significant percentage of women (why women rather than men was never examined) liked to have the TV or radio on most of the day as a source of comforting noise and voices.
I should have gotten suspicious years ago, when I was told by a media research person that Americans watched something like 40 to 50 hours a week of TV. It didn’t quite ring true, and I should have looked for methodological holes back then.
I think the Interactive Advertising Bureau and other industry groups should conduct research on this issue. Rather than puff up the Internet numbers through commensurately sketchy methods, we should create some indexing tools to more properly compare values among media.
We might find that a fair comparison might be a couple thousand TV impressions for each one thousand Internet impressions, or the difference could prove even higher.