The web advertising industry soon recognized cookies as an important tool for targeting users, i.e. for delivering the right ad to the right user. Before cookies, advertisers could reach their target audience only by placing ads on content sites that revolved around a topic related to their product. For example, a computer hardware manufacturer might place ads in the online edition of a computer magazine. This strategy has been successfully used in offline media for decades, and advertisers are always looking for ways to optimize ad placement - like placing an ad for a new printer right next to an article discussing printers.

Through the cookie mechanism, the web opens up a whole new dimension of fine-tuning parameters: By tracking which articles a user reads and which he ignores on a magazine site, the system can compile a profile of his interests. Based on that profile, the system then can deliver not only relevant headlines, but also relevant ads to the user. Instead of targeting broad audiences, advertisers can now target individual users based on their preferences, a strategy known as online preference marketing (OPM).

At first sight, the cookie mechanism's safeguards seem to restrict the reach of OPM to single web sites: Since cookies cannot be read by different servers, a user profile compiled on one site cannot be associated with that user when he visits another site. However, that restriction can be circumvented when both sites use the same server to deliver ads, and the ad server generates and reads all the cookies. The key here is that cookies are not tied to whole web pages, but to individual files; and that web pages residing on one server can contain files from other servers. So, if pages on server A and server B both contain image files (like banner ads) taken from server C, then server C can read and write cookies associated with those image files every time a user views a page on server A or B.

Taking this idea to the real world, A and B would be content publishing sites like online magazines or search engines, and C would be an advertising agency that poses as an intermediary between advertisers and publishers. Since the clients always exchange cookies with the agency's server, regardless of which publishing server they are visiting, the agency can track users across multiple unrelated sites and compile preference profiles with data from all those sites. For example, if a user occasionally clicks on banner ads for scanners when visiting a hardware site, the agency might show him ads touting image processing applications the next time he visits a software site.

The more publishing sites work together with the agency, the greater the synergy effects of this network advertising model: User profiles become more and more detailed, and they can be used by more and more sites at the same time. This is a win-win-win solution for advertisers (because they can target users more precisely), publishers (because they can rely on an existing system with more detailed data than they could provide themselves), agencies (because everybody wants to get on their system) - but what about the users?

< Introduction | PII and Non-PII >