blog.collarity.com

Louise Story wrote a very interesting article in the New York Times this week (To Aim Ads, Web is Keeping Closer Eye on You) about the growing trend of using people’s past online behavior to target them with content and ads. The article infers that the sun may have already set on the anonymous Web 1.0 “On the Internet, nobody knows your a dog” world (Peter Steiner’s iconic New Yorker cartoon), as large Web companies vacuum up behavioral breadcrumbs, appending an ever growing dossier on each of us in an attempt to guess our next move. John Battelle noted the article (That Old Database of Intentions, It Be Growin’), as did David Kaplan (More Behavioral Targeting Than Even Savvy Users Might Expect: Study) at paidcontent.org.

Collarity believes that behavioral data can be applied anonymously, in a way that does not diminish personal privacy. Collarity doesn’t build individual behavior profiles (unless a person pro actively opts-in to this personalization level — normally a very small percentage), which is the main concern of the article. We also don’t track or store IP addresses. Finally, the information we collect is “intra-site” visitor data (we don’t track people moving around the Internet) which is held privately for the exclusive use of our web publisher customers.

Rather than creating person-specific records, Collarity uses data related to people interacting with site content and ads to form groups of anonymous like-minded visitors who are interested in specific subjects or topics. The activity within these implicit attention communities (could be 2, could be 200 on a site) create the headwaters of what we call behavioral relevance. Behavioral relevance is then used as the unified intelligence resource to generate more relevant site search results, provide recommendations (”users who liked this, also liked this”), and to serve advertising that these specific visitor segments respond to most often.

Collarity uses behavioral data (content searching, browsing, ad clicking) more like votes from natural site constituencies. We focus more on what the community-specific vote tallies tell us, rather than the individual voting record or profile of a given community member.

So, at the end of the day, Collarity doesn’t need to know that I’m a dog to be effective. There is no individual targetting of me as a dog or any data element that labels me as a dog. However, if Collarity determines that my behavior correlates with what it understands to be “dog-like” it may harness the knowledge of my clicks to identify which site content is most interesting to dogs or to identify which ads might be most useful for dogs. My behavior may be used, in essence, to normalize content findability and ad receptiveness for other like-minded dogs that arrive on the site after me.