I spent some time today talking to the founders of Fathom Online, a SF- and NY-based search marketing firm. I wanted to get a view of the search economy from a firm that makes its livelihood betwixt and between - optimizing campaigns for advertisers, employing software-based analytics and database mining techniques to eek out the best possible performance from paid search. As many of my readers know, Fathom represents a booming industry, one that will only prosper as performance-based marketing gets more and more complex. I spoke with Chris Churchill, Founder and CEO, Rob Middleton, Executive Vice President New Business, and Jay Webster, President and Chief Technology Officer. Between them they have a boatload of experience in traditional and online marketing, search, and the internet industry.

Our discussion ranged (headline: The Big Agencies Still Don't Get It), but the meme I found most interesting was their take on contextual advertising. I pushed on this piece, both because I feel traditional ad networks fail to take advantage of the endemic value of an audience/author relationship, and because I've heard over and over that AdSense performs miserably compared to AdWords/Overture. But the Fathom guys saw opportunity in my line of questioning. Instead of agreeing that contextual advertising was something of a low-end, low quality option, they said that contextual would benefit from the next wave in marketing spend.

Why? Simply put, the margins on "quality" PPC - search-related keywords on the major engines - are getting too low. The competition is fierce, and there's less money to be made optimizing in those areas. Instead, there's a whole new landscape to navigate, a landscape opened up by AdSense, but which, to date, has yet to benefit from the full attention of the market. But Kanoodle, Quigo, and many others are beginning to play there, offering solutions which, at first blush, are performing far better than AdSense, and for far less cost than AdWords. In fact, Chris said that Quigo was his best performing network last month, so he bought a bunch of inventory there to resell to his clients.

So I then asked the obvious question, at least for me: When would ad networks be able to price both on the keyword/PPC model, as well as price in a "premium" for a particular site, so that an advertiser could say "I'd like to spend up to X dollars to acquire a customer for this particular campaign, but I'll spend X plus 1 dollars if that customer comes from (name your site here - say...Searchblog, for example, or, alright, the New York Times)."

I took comfort in Jay's response: If it can be turned into math, we can do it.

The bottom line: I still predict that within 3-5 years, we'll have ad networks that combine the best of CPA/PPC networks with the best of CPM/Endemic buys. We really are in the early innings.