Table of contents | Next | Previous

Using campaigns

The following article illustrates examples of specific campaigns and recommendations based on customer preferences.

When implementing a personalization solution, it is important to understand the relationship to the content management systems and portals. The IBM Lotus Web Content Management solution uses the IBM Portal Personalization Recommendation Engine, which analyzes Web site user behavior in real time and makes recommendations based on click-through patterns, purchase history, and preference matching. These inputs are fed in turn to the Recommendation Engine subengines, which are subsequently queried for recommendations. The statistics used to drive these recommendations are based on collaborative filtering or Market Basket Analysis dependent on the engines used. If you request a recommendation for the Item Affinity Engine, you receive a recommendation based on Market Basket Analysis. Recommendations for all other engines are based on collaborative filtering algorithms.

The WebSphere Personalization Recommendation Engine is built on the following principles:

• Consumer preferences are not random.
• People who express the same taste in products can recommend products to others. The Recommendation Engine uses collaborative filtering/market basket analysis technology to learn from observed behavior, and based on that behavior, select the right content to present an appropriate product to recommend. At the core of the product is a set of engines that apply the collaborative filtering technology to analyze data .
• Preference Engine: The Preference Engine leverages explicitly stated preferences to make highly accurate recommendations for products and content.
  

The information that is gathered by this process can then be leveraged as rules that can be used to tell Lotus Web Content Management which content, to who, and when.  For our River Bend Web site, we show three examples of this:

1. Campaign Example 1 - In this first example, the user, Dana, is given recommendations of another type of drink based on her buying patterns as shown in the following figure.
   
   
   
   
2. Campaign Example 2 - In the second example, the user, Erasmus, is offered a promotion to get a higher level of discount if he buys a little bit more to raise his level, as shown in the following figure.
   
   
   
   
3. Campaign Example 3 - In the third example, the user Max's profile does not match any of the recommendation criteria. In this case, he gets no recommendations, making this portal experience different from the others as illustrated in the following figure.  
   
   
   

    

   Important : As the profiles of the user change, different portlets are displayed at various times and the portal framework automatically resizes and arranges the portlets so that no space is wasted.