Valleywag notes: Federated Media client losses.
For folks that don’t know, Federated Media is an ad repping firm. If you are a publisher, you can give your inventory to FM and they will sell ads against it. This is different than an ad network because they are transparent that they are selling your ads on your site. An ad network will often stick you into a blind category, like “Tech” or “Automotive” or something similarly broad. Also, rather than paying you a CPM and arbitraging the ads, instead the ad repping firms take 20% (or whatever) right off the top as commissions. Think of them as an alternative to hiring a bunch of ad sales people.
The great part about the FM business model is that they can focus exclusively on higher-quality sites. Rather than working with all the remnant inventory sites that lack focus and have editorially-problematic content, instead they can focus on sites that have branding potential, and thus, deliver higher CPMs. So you’ll see them delivering CPMs in the $5 to $20 range rather than under $1.
The bad part is two-fold: 1) Channel conflict and 2) High/low CPM squeeze
Channel conflict is created once the publisher gets too big. Basically, once the site is successful, then ad agencies will start approaching the site, and the site will want to hire a direct sales team. Of course, this creates competition if an advertiser can buy ads from two different places. So usually what happens is that if the ad sales can be brought in-house, they often are.
The high/low squeeze comes from the fact that ad repping firms can work with sites that are high-quality, but not TOO high-quality. And of course, they can’t work with low-quality sites where automated approaches dominate. In that case, there are just plenty of sites that aren’t high-quality enough, brand-wise, to justify a human (rather than automated) sales team. This is the purview of the remnant ad networks.
Anyway, those are the pluses and minuses. If I have time later on, perhaps I’ll write something about where FM could extend their business model to get around these issues.