ADOTAS – At an abstract level, graph theory is about representing the relationship between nodes (interconnected objects) and edges (the lines that connect them). Applied to the web, the nodes are the pages and the edges are contextual concepts; rendering the web of conceptual relationships across the Internet is a massively complex process, and can make for some really messy images:
Considering its Herculean task, it doesn’t seem surprising that Content Mapper, the technology base of NetSeer’s concept-based contextual advertising, grew out of technology developed by UCLA scientists for studying the complexity of the human genome. And it was this technology that drew online marketing veteran John Mracek to the firm.
“Content Mapper is complicated but intuitive,” says the NetSeer CEO. “I thought, there’s real technology here.”
A computer science major from MIT, Mracek been around just about every tech block, including a stint with Apple during the tumultuous 90s (i.e., before the return of Steve Jobs). Even in the online marketing world, his positions have run the gamut — working on early buy-side media management tool Focal Point; vice president of advertising products at Yahoo!, where he oversaw the development and introduction of SmartAds; and prior to NetSeer, vice president and general manager of distributed commerce at PPC network Shopping.com,
He thinks about Content Mapper this way: When a person views a web page, the brain uses all his/her collected knowledge and experience to analyze what’s on that page. The ultimate goal of NetSeer’s contextual system is to replicate that process and use it serve the most contextually relevant ads or links, depending on the product and customer.
NetSeer’s purely machine-based system with a rudimentary language processor is more scalable and accurate than its competitors’ keyword-based semantic analysis, Mracek says. The mapper, which has more than 70 million concepts and 2 billion relationships under its belt, scours a page for concepts, weighing how they relate to each other, before locating them on the map.
The system recognize “off-page” concepts and weighs the value of conceptual connections — direct, indirect, etc. As a result, ContentMapper builds giant clusters of concepts like the image on top.
Indices of Concepts
Then there’s the business side — first up is ConceptLinks, a tool for publishers that turns display units into lists of sponsored search link depots. Based on the page content, ConceptLinks serves an index of related concepts, Users click through to a search results page (powered by a search engine) that can feature internal links, external links and paid search ads.
The 200 publishers currently using the product have watched search page clicks multiply rapidly. NetSeer comments that the average client page eCPM without ConceptLinks is $3.53; with ConceptLinks that figure goes up to $5.86.
And now it’s going mobile, thanks to a partnership with mobile ad network Mobile Theory. Promising to be a “mobile alternative to Google AdSense,” constantly refreshed indices of concepts will be served to both mobile web-enabled and non-mobile-web enabled sites through Mobile Theory. Business Insider, MinyanVille and Wikia are some of the mobile publishers that will feature ConceptLinks.
The Display Angle
While that’s pretty cool, it’s the Media, or display, product that’s turning heads. After testing in beta last year, NetSeer soft-launched its Media Product at the beginning of the year and focused on garnering business from agencies working with health, travel and CPG clients — Mracek is quick to note that the technology works for all verticals.
Basically, NetSeer scans the various exchanges for content that matches advertiser’s concept list. When a prospective page pops up, it runs through ContentMapper to determine appropriateness; the URL is cached and will have a thumbs up for placement next time it appears.
But what about conflicts, like discount flight ads running next to news stories about plane crashes? NetSeer clients build exclusion lists — Mracek suggests that a dog food company might want to exclude the concept of “Michael Vick” — that fill the need for ad-blocking components.
This keyword-free approach to contextual display targeting already has many success stories, which are likely to grow as more inventory migrates to the exchanges.
While Mracek calls himself a “big believer” in audience buying and targeting, he feels the shortcomings of the sector are coming to light.
“A lot of audience buying is just retargeting, which is probably the most effective method,” he comments. “You may have a great cookie list, but you have to face a lack of inventory. It’s tough to scale.”
But also, without any sense of context, Mracek wonders how much audience buying is leading to wasted impressions. It’s a good question.