December 9, 2009

Is concept-based advertising the future of Internet ads?

VentureBeat

The way Internet ads work right now is fairly basic: Publishers create pages with a certain word or phrase in it, and advertisers buy ads next to those particular keywords. If you search for "guitar," or write a blog about guitars, you're going to see a lot of ads for guitars.

But, wondered NetSeer chief executive John Mracek, what about when you write "bass"? You're going to see ads for guitars, cellos, fish, vocal lessons, and more. And what if someone comes to your site without searching, and there's no data for what they're interested in? That's not good for advertisers, who want the best targeting possible for their ads; and it's not good for publishers, who end up with unrelated and inappropriate ads on their site.

That's why NetSeer is unveiling a totally new way of targeting advertising on the Web: concept-based ads. Instead of tracking keywords on a page as most advertisers do, a process that's prone to error thanks to keyword confusion and SEO-gaming techniques, NetSeer instead looks at the concepts on the page, seeing what a page is about instead of just the words on the page.

The process starts with what Mracek calls NetSeer's "knowledge bank." The bank was constructed by looking through trillions of words on the Web, and finding connections between pages, and between concepts, and mapped the strength of those connections. For "bass," there are a number of connections, but the concept of "bass guitar" is much more narrowly focused, and based on a few different concepts, NetSeer's knowledge bank can understand what a page is essentially about.

After creating a network of more than 50 million concepts, NetSeer came up with a way to figure out, in real-time, what concepts any given Web page has. Thanks to this software, NetSeer is able to target ads much more specifically to the user on a given page, said Mracek. NetSeer's system doesn't know that bass is a type of fish, but knows that it's closely connected on the Web to bass fishing, tackle, and fresh water fish. That's all data advertisers kill for.

The benefit of all this to advertisers is that they get to place much more targeted, and often cheaper, ads on sites where people look for products like theirs. One example Mracek gave was migraine headaches: migraine headaches, in NetSeer's knowledge bank, turned up, among other things, a connection to "dried fruit." One of the ingredients in dried fruit, he found out, can cause migraine headaches-by telling advertisers about dried fruit sites to buy ads on, companies are able to target their audience in more relevant places, and might be able to avoid buying ads on any page that says, "I had a migraine today."

Instead of coding information about the user, like what they search for or where they come from, instead NetSeer codes the page itself, determining what the page is about, and can target ads assuming that people come to the page for what it's about, rather than relying on search data. The result, says Mracek, is higher click through rates for advertisers, better CPM for publishers, and a more targeted and useful ad experience for users. If you're on a diabetes page, regardless of what you searched for 20 minutes or 3 days ago-as many sites, like Google, track to figure out how to advertise to you-NetSeer will show you ads about what you're looking at.

There are already about a dozen publishers using the technology-like i4u, whose "Sponsored Links" come from NetSeer-and the results are impressive: in a "Video Game Hardware" article, I saw ads for the Nintendo Wii, the Playstation 3, and memory sticks. For a look at how concepts are related, NetSeer has a concept demo site-just plug in a URL, and NetSeer's knowledge bank comes up with related concepts for you to see.

The company plans to roll out this technology two-fold: they will both sell and target ad space themselves, and license the knowledge bank to other companies looking to take advantage of concept-based advertising.

NetSeer, a 23 employee company founded in 2006 in Santa Clara, Calif., has already raised $14 million in funding.