Behavioral Targeting in Online Advertising

April 18, 2009 by
Filed under: Uncategorized 

In my new job as a data mining consultant at FinScore, I’m currently working on behavioral targeting for online advertising. The client is a telecom company in Switzerland. The main idea behind the project is to show ads on their website that are dependent on the profile of the visitor. In theory, online advertising is much more powerful than radio or TV advertising for example. This is mainly due to the lack of possible “targeting” in such medias.

This targeting is quite trendy these days since it is part of the foundation of the Web 3.0, which means a web based on personalization and recommendations. Google recently started to use behavioral targeting for AdSense. Here are more details about how Google targeting works.

The first part of the project consists in aggregating ad impressions and clicks from web logs. This newly created data base will be mined to build a behavioral model for each distinct user. This model is applied to visitors who didn’t see ads in order to decide which ads to show them. The final objective is to increase the CTR (Click Through Rate).

For the technical part, I use SAS tools. As I had already discussed on Tweeter, SAS tools such as DI Studio, are not easy to use and their GUI is quite old. SAS Base language, although powerful, is not as straightforward as R for example. However, the SAS support is really good and their answer time is usually less than a day. For the web mining part, I have read the book Mining the Web: Transforming Customer Data into Customer Value (I will soon write a small book review about it). For the SAS part, I’m reading the excellent The Little SAS Book (I really recommend it to you, if you’re a SAS user).

For more information about online advertising and behavioral targeting.

how tall will my baby beWhen it comes to channels, it’s somewhat of a toss up between the two giants. Each service has hundreds of channels available, including local channels available in each and Cheap Football Jerseys every package. Direct TV has dominated the niche of sports programming, so if you’re a big sports fan then you’ll probably want to go with Direct TV. Remove the steak and allow it to “rest” on a flat surface, 10 minutes. Cooks Note: If you like your steak a little more cooked, simply leave it in the skillet a few minutes longer on each side. Wipe out the skillets and use them to warm the tortillas, 1 at a time, over low heat. Hire a strength and conditioning coach. You will want to ensure that you are in the best shape that you can be in for when you try out. This person will put you through the paces you need in order to cheap football jerseys compete on the same level as others at the tryouts. ZEIGLER: Well, there were a couple of things. First, as far as before or after the draft, Michael wanted to go to a team that wanted him. And he had been out to his team and he did not feel like going back in the closet. With that, we set off together to find this dick. Only Cheap Football Jerseys 9,000 Nazis escaped to South America, but the SS, who did most of the dirty work of the Holocaust, numbered 800,000 at its peak. Most of them tended to stay home when they were done war criming. I’m going to show you a variation that I do. V step facing forward and a V step facing the back of the room, which is off the bench, a little L with a squat, a revolve, and then my custom jerseys hamstring curls. Since we did a lot of movement off to the side and around the bench, I like to come back with kind of a stable movement in the center, just a corner to corner, rather than adding anything fancy after that. 10, 2015. The agency’s latest count includes 39 emergency room treated injuries that are fall and collision related. The hospital visits included 16 fractures, seven strains or sprains, five contusions or abrasions, two concussions and two additional organ injuries to the head. Kilpatrick, a 39 year old physical education teacher, believes racism and limited opportunity are the Wholesale Jerseys reasons behind the controversy. “Whether you agree or disagree, Caucasians have been known to obtain jobs based solely on their skin color, with Blacks it’s just the opposite,” she says. Kilpatrick noted that although it may be viewed as reverse racism, many believe that Blacks are athletically superior to Whites. Gaming is one of those entertainment industries that is often called “Recession Resistant.” In tougher times, the theory goes, people need their escapism and fun more than ever. Even as they watch their spending, they’ll pay for a movie ticket, or a DVD, or a new game to suspend reality. But most anyone who has ever cheap oakley sunglasses owned a cheap nfl jerseys water resistant watch will tell you, water resistance isn’t the same as waterproof.


11 Comments on Behavioral Targeting in Online Advertising

  1. steffen on Wed, 22nd Apr 2009 7:44 am
  2. Hello Sandro

    Thank you for this nice comprehensive article, especially the technical terms, which sometimes are hard to identify. Not its easier to dig deeper. Also thanks for the book link. It looks very interesting. I hope you’ll soon have the time for a review.

    kind regards,


  3. Sandro Saitta on Sun, 26th Apr 2009 1:03 pm
  4. Dear Steffen,

    Thanks for your comment. I will write my review soon.


  5. Ahmad on Sat, 20th Jun 2009 7:45 am
  6. Hello Sandro:

    With regards to behavioral targeting in online advertising, do you have any idea of how to get online advertising dataset available for research purposes?


  7. Sandro Saitta on Sat, 20th Jun 2009 3:54 pm
  8. @Ahmad: I think this kind of data (user web profile) are crucial for companies (for ads targeting, personalization, etc.). I think it would be very difficult to have (buy?) them. However, I have read some research papers of people who have worked with a search engine in particular.

  9. Ahmad on Sat, 20th Jun 2009 5:08 pm
  10. Hello Sandro:

    Thanks for your reply. I understand that some people managed to work with the advertising dataset. These people mostly work with search engines, and naturally the advertising dataset is available to them since search engines gain from ad display. However, the dataset is generally not available to the public.


  11. Nicolas on Wed, 18th Nov 2009 10:27 am
  12. Very good post and very good links, thanks.
    I am working on the same line of research (behavioral targeting), but i´m using Ontologies in my research. Do you know any network that connects user profiles and categories of ads?. Thank you.

    A greeting.

  13. Sandro Saitta on Thu, 19th Nov 2009 8:40 am
  14. @Nicolas: Thanks for the comment Nicolas. I don’t know exactly, but I think United Internet ( has something called Target Group Planning. It may be what you’re looking for. Unfortunately I have only found a german page. Otherwise, you can try Wunderloop for example. They are linking user profile in terms of demographic data to some groups of ads (i.e. kinds of interest groups).

    BTW, if you have more information about your project I would be interested.

  15. Nicolas on Sun, 22nd Nov 2009 8:09 pm
  16. Hello Sandro,
    thanks for your links. I was looking at websites dijists that you gave me, and although I didn´t find the information, them have been interesting. I’m working with the Protege tool among others, and i try to make with ontologies to segmenting users and ads. Know something of ontology?

  17. Sandro Saitta on Mon, 23rd Nov 2009 9:45 pm
  18. @Nicolas: I just know the basic concepts of ontologies. What do you mean by segmenting users and ads? We can continue our discussion by e-mail if you want:

  19. John on Mon, 20th Jun 2011 5:07 pm
  20. Sandro,
    Great article, though I see it is more than 2 years old, I wonder if this topic is still interesting to you? (It is to me!).

    I have a specific question-it seems that behavioral targeting (based purely on web behavior) and demographic targeting (based on known customer attributes) are conceptually very powerful concepts to link together. However, in practice, they are almost ‘oil and water’ and seldom combined. I wonder if your previous project gave you any insights into the power of linking the two?

    Second, I am looking for a practical ‘cookbook’ for how to approach this type of problem with SAS or SPSS (in terms of CTR modeling) and wonder if you have any book/article recomendations.

  21. Sandro Saitta on Tue, 21st Jun 2011 7:57 pm
  22. Dear John,

    Thanks for your comment. Although I don’t work in BT any more, I still find the area very promising.

    Regarding your first question, this is one of the particularities of the tool FinWEB developed at FinScore (I was working on this tool).

    About the cookbook, I don’t know any book about SAS and CTR issues. However, “Mining the web” can be a good start. Feel free to contact me by email to continue the discussion (I can then point you to the right person).


Tell me what you're thinking...

  • Swiss Association for Analytics

  • Most Popular Posts

  • T-shirts, Mugs & Mousepads

    All benefits given to a charity association
  • Data Mining Search Engine

    Supported by AnalyticBridge

  • Archives

  • Reading Recommandations