Google at the Crime Scene.

Google at the Crime Scene: Profiling Websites, Estimating Traffic, and Combating Click Fraud.

If you own a web site, Google wants your traffic details, or at least they want to be able to estimate your traffic and the behavior of visitors at your web site. It’s something that might help them uncover click fraud by discovering unusual amounts of traffic and unusual behavior by visitors to sites.

They’ve been handing out tools to let people measure their own traffic, such as Google Analytics.

They’ve also recently announced the release of Website Optimizer (beta), which is a program that lets Adwords users test different variations of their landing pages. It appears that between the two programs, and some reasonable guesses, Google might be able to use the information provided to help build traffic models that might be used to help them combat illegal clicks.

Profiling Web Sites

One of the initial steps in estimating traffic to pages is profiling those sites. Here are some things that they might include in this profile:

* The primary Language used by the site (English, French, etc.)
* A category for the site by type or intent, such as a shopping site or news site
* A category for the site by industry classification (targeted towards computer enthusiasts or home contractors, for example)
* The geographical location of the site’s host
* The average price of a product for sale (for shopping sites), or;
* Information about the HTML layout of the site (mostly text, or graphics, etc.).

Traffic to different sites may be estimated upon these profiles, and detailed statistics about the sites. If there are unusual traffic patterns uncovered at sites that display Google’s Adwords, that may be a sign of foul play (click fraud).

Estimating Traffic Based upon a Traffic Prediction Model

When Google knows about the traffic to some sites that fit a particular profile, they can estimate the traffic to other sites. A Traffic Prediction Model may also include values or tags describing:

* Expected traffic, including the behavior of the traffic, at a site.
* How long a typical user will spend at the site,
* How “deep” or how many links a typical user will click on within the site,
* The screen size of a typical user,
* The type of browser of a typical user, and;
* Demographic information of a typical user.

When someone uses Adwords, Google might use a traffic profile to estimate the amount and kind of traffic to their site. The initial traffic profile could be based on a known traffic profile of traffic at another site that has a similar set of descriptive tags as the initial set of descriptive tags determined for advertiser.

Different kinds of sites have different traffic profiles. For instance, at news site visitors tend to focus upon specific articles and read through them until they have finished. Traffic at a product comparison site usually sees behavior from users who are more likely to stop reading a particular product page once they find a product that meets their needs.

Traffic that doesn’t match the profile might be suspect.