mengdan

Slide 1: 

Mining E-Business Gold Web mining application to customer relationship management(CRM) Dr. B. Berendt / Web Mining Seminar Humboldt University Mengdan Yu Jan. 2002 

Overview : 

Overview  Data source  Task & algorithm of Webmining application by tracking CLC  Conclusion

Slide 3: 

Data transformation cycle New Question Data Knowledge Decision Information $ $ $ Turning data into dollars is a continual process of asking questions of the data, extracting information, transforming it into knowledge, making a decision, and taking action in hopes of producing a significant customer retention - and then starting the process all over again. Figure 1

Slide 4: 

Knowledge The process of extracting knowledge from data is called Web mining . Data

Slide 5: 

The new challenges of web-based CRM “Data rich” v.s. “Information poor” How to maintain real-time contact “Passive” v.s. “Active” How to deal with newly-empowered customer “One fits all” v.s. “one to one” How to enhance customer intimacy “ Six” v.s. “6, 000” How to leverage network value

From Web log to Web Loyalty : 

From Web log to Web Loyalty A study done by the Harvard Business School indicates that an increase of 5% in customer loyalty can increase profitability from 25% to a much as 80%. (Multimedia Live, 2001)

Slide 7: 

The Customer Life Cycle Figure 2-a Detection and prediction

Slide 8: 

Four Ts of CRM in the Web Targeting: market segmentation, that is fine tuning and aligning one’s sights on the target public. Tailoring: the adaptation of the offer to suit the buyer Typing: Linking up with the customer, to establish a long-term relationship based on the trust and satisfaction. Tapping: The valuable source companies benefit can be tapped to give further profits by using the customers as references for prospective buyers.

Slide 9: 

Source of Data  Subscriber/membership analysis occupation, income, age, geography, etc  Website content, usage and structure analysis IP address, user ID, time/date stamp,cookies, etc Three levels: server/client/proxy  Customer behavior analysis clickstream, pageview, clickthroughs

Slide 10: 

Web ming application by tracking CLC  Identification reach and acquisition stages of CLC  Detection converstion stage of CLC  Prediction retention stage of CLC  Optimization loyalty stage of CLC

Slide 11: 

The Customer Life Cycle Reach Acquisition Conversion Retention Loyalty Abandonment Attrition Churn Reactivation Figure 2-b Identification Detection Prediction Optimization

Slide 12: 

Identification -- stages of data quality control Figure 3

Detection : 

Detection Statistical Analysis - Descriptive statistical analyses(frequency, mean, median) can be applied to analysize session files on variables such as page views, viewing time and length of a navigational path. Cluster detection - to group together a set of items having similar characteristics. These clumps of self-similarity are called clusters. Decision trees - It divides the records in the training set into disjoint subsets, each of which is described by a simple rule on one or more fields.

Slide 14: 

Predication Dependency modeling - to develop a model capable of representing significant dependencies among the various variables in the Web domain. Sequential patterns - to find inter-session patterns such that the presence of a set of items is followed by another item in a time-ordered set of sessions or episodes. Association rules - It refers to sets of pages that are accessed together with a support value exceeding some specified threshold.

Slide 15: 

Predication Memory based reasoning - looks for the nearest neighbors in the known instances and combines their values to assign prediction values. Artificial neural networks - They are simple models of neural interconnections in brains, adapted for use on digital computers. In their most common incarnation, they learn from a training set, generalizing patterns inside it for prediction. ( Cont’d )

Slide 16: 

Optimization  Self-adaptive Websites automatically improve their organization and presentation by learning from the Website user’s online behavior based upon certain pattern revealing techniques  Personalized recommender system  Neighbor-based collaborative filtering  Singular Value Decomposition  Regression-based approach

Slide 17: 

Conclusion  A more statisfactory customer, a more profitable organization  CRM (three P): Philosophy, Process, Perception  Tracking CLC  From data to knowledge: Webmining technology to improve data accessibility and operational efficiencies

Slide 18: 

 Open questions: dynamic data source real-time customer support electronic interaction v.s. human contact Conclusion ( Cont’d )