InsideGoogle

Wandering Around The Exhibit Floor

I walked around the exhibit floor, taking a look at the various search engines and ad companies. Here are a few things I saw:

I met with Tom McGovern from Snap, the search engine by the former Overture team, and I continue to be impressed by these guys. Snap just introduced search advertising, something they call “Cost-Per-Action” advertising. While most everyone else offers pay-per-click, where the advertiser only pays when the ad is clicked, Snap offers the option of only paying out when a specific action (chosen from an appropriate list) occurs. As long as the action earns the company more than the ad cost, this represents the perfect risk-free advertising.

Thing about it this way: Staples buys an ad for 3% on searches for binders. Searcher clicks on ad, buys nothing, Staples pays nothing. Searcher buys a single binder for 75 cents, Staples pays 2 cents. Searcher turns out to be a company, buys 5,000 binders for $3,750, Staples pays Snap about $112.

If Staples wants to be smart about it, they can buy cost-per-action ads at variable prices for regular searches, so they pay a percentage-based comission per sale, and fixed prices for searches on business keywords. This would result in a 2 cent payment on a single 75 cent sale, and if you pick your business keywords properly, sales resulting from those terms could be 40 cents, instead of $112.

See, by mixing variable pricing and fixed pricing, you could pay tiny commissions on small sales, and fixed costs on bigger ones, just by picking the right keywords. The best part? You pay nothing if they buy nothing. And of course, you can always use traditional pay-per-click if you’d like.

One thing that struck me as odd is that I was told the rankings in Snap’s search engine are based partially on the ads. While this isn’t exactly paid search, their relevancy algorithm takes into account the success of the ads. Snap tracks how many people who click on the ads convert into sales. It not only uses that data for billing, it also uses it to determine relevancy.

In Snap’s eyes, an ad with a high conversion rate represents a site with highly relevant data. As a result, that site will rank higher in the unpaid listings, representing a version of paid search that isn’t exactly paid for. What Snap’s advertisers get is higher search rankings based on ad performance, a major advantage over non-advertisers. It might be highly useful for a company to buy a dollar of advertising on Snap, make sure its highly relevant and cannot possibly not result in a successful action (such as watching a video), just to get the effect of paid search with higher search rankings.

Call it “backdoor paid search”.

Tom also told me a “fun factoid”. Snap’s corporate name is Perfect Market Technologies. Why? It’s a reference to Adam Smith, who laid the framework for the free market, arguing for the very transparency Snap is known for. See, not only does Snap offer some incredible transparency in the form of revealing all of its site stats, Snap’s new ad program takes it even further by telling you how much the ad is costing the advertiser, and how many people clicked on it! Unbelievable.

Who else did I talk to? Well, I went to Yahoo and got my ticket for tonight’s party. I also went to MSN Search, and found out they were holding a party tomorrow night, and got an invite to that as well, plus an MSN Search T-shirt to go with my SES New York shirt. Also, did you know MSN Search can solve complex equations? Just enter 4x+5=12 into the search box, and let MSN solve for X.

I also got to sit down with Dmitry Eroshenko, CEO of Clicklab, who’s company has just announced a new direction: detecting click fraud. Clicklab’s business used to be website analytics, so they think that puts them in a unique position to be able to analyze site data and detect frauds. They look at all the site log data, with Javascript tags inserted on every page, tracking every user that arrives via an ad and comparing it against a myriad of data. The system then calculates a penalty score, and if its too high, a report is generated notifying the site owner that a discrepancy has been detected. There are multiple tests, and a single fail generates a report.

Clicklab has been testing this system for two years and expect it to perform very well. They can’t publish fraud data for individual sites due to NDAs, but plan to issue reports of how much fraud is prevalent in different categories, and I will have those reports to show you their progress. Because site owners do not want to give their detailed site info to Google and Overture, Clicklab hopes to position itself as an independent party to the ad delivery companies, one they can trust to accurately report click fraud data. They have already initiated talks with Overture/Yahoo, and hope to bring Google aboard as well.

Last thing, I finally got to see a Google Cheese Box in person (that’s the Enterprise Search Appliance to you), and boy is it big. Their enterprise person had ditched the booth, so I’m hoping to get a demo tomorrow. I did walk away with a Google Desktop Search notepad and the little Google fish you can see perched on the Cheese Box.

March 1st, 2005 Posted by Nathan Weinberg | General | 4 comments