A fascinating new report from Nielsen/NetRatings is claiming the unthinkable: that the major search engines are running out of affordable ad inventory, with demand far outstripping supply. Whoda thunk we could have reached that point so quickly?

The core issue is this: the search engine model is to bid on keywords, and though Google is proliferating through to thousands of other Web sites via its AdWords program, an advertiser who buys a single keyword will appear on all relevant Web sites. There's no further selection or targeting capability, and that limits supply. There may be an infinite number of words, but there are apparently a very finite number of words people are willing to pay to be associated with, and a limit to how much they are willing to pay to be associated with them.

Since these keyword programs are priced using an auction model, with the search engine making its money based on the number of click-throughs, it could be that the big search engines have painted themselves into a corner, albeit a profitable one. That's because the current scenario will mean advertisers will have to pay increasing amounts to maintain the same level of visibility. At the same time, however, it doesn't appear to be in the interest of the search engines to create more targeting options for advertisers, since they make more money for a lot less work in the current model. Indeed, prices could potentially drop if the search engines created more targeting options, since that would effectively create more inventory.

The other fascinating trend is the rapid growth of software products and service firms dedicated to managing paid search programs, and they all seem to be flourishing, given the inherent complexity of the auction process, and the real need to constantly tweak both keywords and bids to achieve maximum results without paying too much. There really is no precedent for such a high level of involvement by advertisers in overseeing their own advertising programs. As one example, Overture has just released an application called Search Optimizer that sits on top of its existing Marketing Console application, and is designed to ease management of large numbers of keywords. Rather than offering the product free to big customers to mitigate the complexity of its own system, Overture is reportedly charging an average of $500 per month for this service. Ouch.

Put together these two things: increased costs and increased complexity, and what do you get? Advertiser backlash. No, we don't think paid search is going to collapse. Indeed, the Nielsen/NetRating reports suggests to advertisers that for now the results are still worth the cost and the hassle. But with affordable keyword inventory drying up, some advertisers may be forced out of the paid search market.

What's it all mean? We think the stage is being set for a prediction we first made a year ago: a two-tier search market, with B2B buying guides purchasing keywords to drive traffic to their sites, and then selling space on their sites to advertisers, effectively reselling the keywords.

Why will this happen? As popular keywords become increasingly expensive, among the few who can afford the cost will be those who directly or indirectly are reselling them. Further, in this two-tier arrangement, advertisers get the benefits of keyword marketing without the oversight hassles, and effectively share the cost. The publisher then assumes the role of buying the proper keywords, managing the program and paying for it, while the advertiser goes back to the traditional and far easier role of simply writing a check.

Another great benefit to this approach, one that publishers and advertisers are just beginning to recognize, is that visitors that come to them this way are in effect "triple filtered." It's highly unlikely that someone who clicks on a keyword driven contextual ad, then visits a vertical B2B site, then clicks on an individual advertiser link is doing so accidentally. That means better quality clicks, and more value to the advertiser.

We've long maintained that the big search engines couldn't succeed by trying to be all things to all people, but we admit we never anticipated that they would become victims of their own success.