The Art And Science Of Online Profiling

Last week, we touched on the topic of online profiling. But this week, let’s roll up our sleeves and really get down and dirty with profiling techniques online marketers can use to collect valuable information about their customers.

Online profiling is as much an art as a science, and it is at the heart of any targeted or one-to-one web marketing initiative. At its core, it blends the creativity of marketing with the technology of database applications.

The Science

There are many online profiling and personalization systems available, and each uses its own database logic and other methods such as rules-based collaborative filtering, statistical modeling, and intelligent agents. But here’s a basic outline:

Step 1: A customer fills out the registration profile web form. These forms typically contain specific questions the user answers using checkbox interest selections from a pre-designated list of choices. Other customer registration information such as name, address, and so on, is captured via entry boxes.

Step 2: The profile information is added to the profile database, tagged with a unique ID using a cookie stored on the user’s computer, or with a bookmark containing their unique ID, or with a password-protected web page where the user enters an ID and a password each time he or she visits the web site.

Step 3: The web site presents customized or personalized information based on an initial profile. The web page may not be generated until the profile has been entered, so it can be populated with information that is associated with the user’s profile. Many times, however, a “guest” profile is used with standard, or default, settings. The “template” web page used by the personalization software to create the personalized page seen by the user contains special tags that instruct the database to retrieve unique information, populate the template, and then display the web page. A wide variety of information, product recommendations, advertising, email and so on, can be personalized in this process. These web pages are considered “dynamic” rather than static.

Step 4: The customer browses the web site, and the system tracks the user’s clicks and adds this data to the web server log database. This “click stream” data can be used to automatically alter each individual’s profile data in much the same way that the form data supplied by the user updates their own unique profile.

Step 5: The customer returns to the web site, where the site is displayed in a personalized manner based on that user’s individual profile. Further user interactions are logged by the database, such as user answers to additional questions about how they rate certain content or products, or information on purchases made, and sometimes additional click stream data. Web pages are personalized through the use of “rules” (for example, if a visitor lists scuba diving as a hobby, then a scuba promo may be inserted), or by using a variety of complex statistical formulas.

Step 6: Repeat visits over time will provide additional information to help you understand and serve your users or customers better. By joining the profile database with the web server’s log database, publishers can answer a wide variety of marketing questions, such as, “How many scuba divers saw article number seven?” Or maybe, “Is order size related to the number of times visitors stop by our site? Or is it related to the time they spend on our site?”

And The Art

The art of online profiling revolves around one important asset: Your customers. Or as my co-author Cliff Allen often suggests, “You need to think like a customer.”

To build an online profile database, you need to know what information is important to ask for, based on your knowledge of your customer. Of course, only ask for that information critical for one-to-one web marketing all else is superfluous.

Collecting too much information is a waste of resources, certainly. But it also threatens to create distrust among customers, who may be wary of providing too much or too detailed information.

How does the marketer think like a customer? Start by asking yourself these questions: How do our customers interact with us? What are their interests? How do they make buying decisions?

Since most consumers make decisions based on a small number of factors — probably less than six — it only takes a few well-chosen questions to gather a tremendous amount of data about each individual.

For example, if you have only six checkboxes on your profile form, there are 64 unique profiles possible — more than most of us now use in tailoring our sales presentations! Of course, with multiple-choice questions that use radio buttons, it’s possible to have a very refined profile of interests.

“By using a decision-tree approach to analyzing customer purchase behavior, it’s possible to plan a personalized web site that uses a small number of online profile data items to recommend the best products to your online customers,” advises Cliff Allen, a web personalization consultant and co-author of the Internet World Guide to One-To-One Web Marketing.

As I indicated last week, online profiling can be performed using declared and/or behavioral data. To see how online profiling can work with real customers, take a peek at the below. Observe how they build profiles — the types of questions they ask, the sort of information they garner. (You’ll notice all of the sites listed use cookie technology.) Each provides some personalized information or services that are based on online profiling.

  •’s Recommendation Center
  •’s Personal Framing Service
  • BabyCenter’s Personal Page
  • Backyard Nature By Design
  •’s Personalized Page
  • Garden Escape’s My Stuff
  • Purina’s Recommendation for Dog Breeds
  • Total E’s eDNA Personal Shopper (Columbia House)
  • Travelocity’s Personal Profile and Fare Watcher Email

    Next Week: Time to talk about that issue that makes many marketers squirm: Online privacy.

    Related reading

  • <