Another point to support your argument is that in the current web ecosystem, there are few (if any) individual publishers that have the necessary reach of specific audiences desired by advertisers.

All excellent points.

Using "house ads" to drive traffic from low value pages to high value ones is a tough game and the economics rarely work out. Many sites use links and other tools in the traditional editorial areas areas of the site to do just this. Giving up the value of display impressions to try and move clicks to high value areas typically yields very low ECPMS for the publisher.

Using BT to extend audiences in areas that are oversold is exactly what Tacoda and Audience Science do for publishers. This tactic works quite well but is very difficult to do at any scale, especially for smaller pubs.

And I agree. If you are a publisher with very high sell out rates, there is not need for a network partner.

Andrew

Not much point for a premium publisher that has decent sell-through to work with a network. Exchanges would seem to be a logical though.

Different story for publishers with low sell through, they have to work with networks...

I wanted to add that proper use and segmentation of user data can solve these problems. The publisher always has the power to carve out these premium impressions. But having the proper data available to Ad Networks can turn remnant inventory into Premium. And having relevant ads from BT and RTB campaigns solves the performance problem.