Project overview
Recommender systems are becoming increasingly important as they can help businesses to offer more personalized product recommendations to customers. These recommendations are usually based either on what the customers have already bought, or on what “similar” customers have purchased. A driving principle behind recommender systems is: instead of the user seeking for information, the appropriate information is “finding” the user.
The power of recommender systems is based on the effective utilization of the voluminous data, in order to “fill in the blanks”, thus enabling more accurate prediction mechanisms typically in the form of “what will the customer buy or prefer in the immediate future?”. In our research and development activities, we offer a different perspective by addressing the problem not from the customer side (“what can the system offer to me?”), but from the sales and marketing viewpoint (“what can we offer to the customer given all the information available?”). Thus our effort is different from traditional recommender systems (such as Amazon, or Netflix), which have limited knowledge of their customers, but a more complete knowledge of products and customers' buying patterns.
In a current project with IBM Switzerland, the customer base represents a large number of small and medium-sized companies, about which information is collected from diverse sources, including financial reports, news events on the Internet, etc. Our aim is to build more realistic customer models by examining not only their buying patterns, but also by evaluating information that spans almost every facet of their operations. Therefore, we hope to offer a more holistic view of recommender systems, allowing for a closer analysis and monitoring of the joint customers–products portfolio. In that sense, our effort represents a mesh between graph mining approaches and recommender systems analytics, while touching upon numerous other research areas including information retrieval, data representation and visualization, and data mining.
Example of overview of various customer data views offered to the user. Information provided includes: (a) customer firmographics, (b) contact people, (c) open and closed sales opportunities, (d) blog data, and (e) RSS news feeds.