VA

Rules-Based Filtering

Rules-based personalization generates a profile of each customer, which is stored in a database and used to identify patterns of behavior. The patterns are transformed into assumptions, or rules, which are then used to predict a shopper's future likes and dislikes. Then, the e-tailer can customize his or her content accordingly and tailor sales and advertising efforts.

There are several ways to accumulate this data. The most basic collection process is the use of a registration form, which requires visitors or shoppers to fill out a questionnaire and submit it to the store or Web site before they can shop. Requested information might include age, occupation, address, household income level, hobbies, interests, and so on. Most e-tailers, however, have abandoned the practice of registration in favor of a more subtle approach, which is less likely to scare off potential customers.

One of the more common strategies you see today is the use of clickstreams, which track the path that users take through a Web site. Most of these types of programs allow companies to track the number of times viewers click on advertising banners, giving them the data they need to customize advertising campaigns to their viewers' preferences. This practice is similar to what traditional retailers do when they log in purchases for future direct-marketing campaigns. Another increasingly popular tool is the use of if/then scenarios, in which e-tailers attempt to direct the movements of users by basing the user's options on what they've done in the past.

Collaborative Filtering

This is the process by which retailers and e-tailers track customers' likes and dislikes and look for patterns similar to other customers. This concept is designed to simulate a "word-of-mouth" campaign.

Personalization isn't just looking at one person's behavior. It's looking at the behavior of people who are alike. And it's making a set of suggestions based on what a particular group of similar people think is popular.

Online music store CD Now is one site that implements an advanced collaborative-filtering process. CD Now takes information about what you look at and associates you with other people that have similar preferences to you, then, it says, 'Other people who look at those kinds of sites have purchased and listened to this album and rated it highly, so we're going to recommend it to this person.' It might be completely off the wall, but there's still a high likelihood this person will buy it.

The more times a person visits and makes purchases from your site or store, the more likely you will be to make helpful purchasing suggestions to that person.

Dell Computers

Dell, one of the most successful online computer retailers on the Internet, has designed its personalization tools to make it easier for shoppers to customize their own product specifications. The system guides customers through a range of choices in RAM, hard drives and peripherals, helping them avoid compatibility conflicts along the way. This is a case of not overdoing it with personalization and increasing serendipity. Dell doesn't waste users' time with needless features; instead, it personalizes only what the user needs to get a good computer. At the same time, the same referencing tool that steers the customers to compatible products also steers them to products they might not know about, such as different manufacturers' equipment, etc.

Clinique.com

This online cosmetic retail site offers personal consultation based on a series of interactive questionnaires visitors answer. Product recommendations are based on the skin type, color of hair and eyes, pore size, breakout tendency, etc. of each visitor. The site remembers the answers so each time a visitor returns and enters her registration information, her information is on file.

Food.com

This online dining site provides a great example of seamless and unnoticeable personalization. CyberMeals uses Apple's WebObjects technology to automatically give users information on food delivery or take-out restaurants that fit into their tastes, schedules and whereabouts. Much of this is taken from the user's ZIP code information and past orders.