“We need to find ways to target people of particular demographics that are comparable to the people you might find in The New York Times or a particular publication that you may be familiar with.”
That's a senior VP from Google interviewed in the Times (dug up by Rough Type's Nick Carr), talking about the search giant's shortcomings in selling ads on social networks. That's the trick about turning your ad sales over to an algorithm--there's no good way to target demographics and, as a result, for every spot-on ad that gets placed, there are plenty of total misses (like the ones in the sidebar on this page, which are currently advertising get-rich-quick schemes because a few posts on the front page deal with the challenges of making money with online publications). In the social networking space, where most pages aren't going to return easily computed subjects (how do you come up with an ad for a page filled with people commenting "thanks for the add"?) the challenges are even more daunting.
One wonders if that's part of the driving force behind Google's new initiative to create a semantic social web: Once you can define relationships between users across sites, you're starting to create a machine-definable demographic.
Until then, though, Google will be looking back fondly at the ease of how a publication could define its readers and then sell them to high-value advertisers. Ironic then, that those very same publications now look to Google to bail them out of web advertising.