How Behavioral Targeting Works

Most implementations of behavioral technology work by tracking a person by setting a cookie on his computer. As he moves across the website (in case of on-site targeting) or different websites (in case of ad network targeting), with every request from his browser this cookie data is passed onto server responsible for behavioral targeting.

This server records a host of (anonymous) data. Typically, following types of visitor specific data can be tracked by the server (note that the list is not exhaustive):

• City and Country from where the visitor is accessing the website (useful for geo-targeting)
• Local time of the visitor (useful for determining if the visitor is a night or morning person)
• Browser used, Operating System, Screen Resolution, Internet Speed (is visitor accessing the site from his mobile? Is visitor technologically more savvy and using Linux? It is amazing how much information do visitor system statistics reveal.)
• Type of  page viewed by the visitor (made possible by tagging the page and analyzing it by the server)
• Time spent on page, when did he accessed this page last time and number of times accessed the page (tells us about interest, recency and frequency of the visitor)
• Made any purchases? Value of those purchases? (gives us an indication of how monetarily valuable a visitor is)
• Where did this visitor come from? If search engines, which keywords did he use? (useful for targeting content based on the inferred intent through the use of keywords –for example, for the term “buy” and for the term “free”, you would want to tailor the content differently)

All this information, along with the information previously tracked as visitor moved across pages, is used to make a profile of the visitor. A visitor is said to belong to a particular profile (or segment or bucket), if he satisfies a criteria associated with that profile. For example, when a visitor freshly arrives on a music site, he may be put into “New User” profile by default. Now, if the visitor makes a purchase in Guitar section, he may be put into two profiles “Buyer” and “Guitar-Interested”. As the visitor makes another visit after say a week, he is put into another profile “Repeat Visitor”.

When we plan to display content or advertisements to the visitor, we look at what profiles does the visitor belong to, how recently was he allotted to those profiles and make a decision accordingly. For example, there might be following business rules:

• Cross-sell guitar lessons to the visitor who belongs to segments: “Repeat Visitor”, “Guitar-Interested”, “Buyer”.
• Offer a discount to the visitor who belongs to “Repeat Visitor” but not “Buyer”
• Promote free online guitar lessons to the visitor who belongs to “Guitar-Interested” segment

Behavioral targeting can accommodate the rules described above with ease. In fact, even more complex targeting rules are possible: including tens of segments, which further includes tens of criteria. It is important to note that, all this processing takes place on-the-fly. As visitor browses the website, his profile is assembled plus he is being targeted at the same time using the amount of information the system currently has.