2006bai6486.doc.doc

209 views

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
209
On SlideShare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
4
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

2006bai6486.doc.doc

  1. 1. A DATA MINING CASE STUDY IN THE UNDERWEAR INDUSTRY FOR CRM APPLICATIONS Ta-Wei Hung, Department of Information Management, Shih Chien University, 70, Ta-Chi Street, Taipei City 10462, Taiwan ROC twhung@mail.usc.edu.tw Jing-Yin Kang, Institute of Management of Internet Trade, Shih Chien University, 70, Ta-Chi Street, Taipei City 10462, Taiwan ROC 8822030@pchome.com.tw ABSTRACT This is a case study in applications of customer relationship management (CRM) for company C that manufactures and sells the female underwear as its major business. Among many brands of company C, we only study in applications of CRM for one brand named brand B. Therefore, the database regarding brand B transactions is studied for CRM applications. Specifically, the table of transactions data, the table of transactions detailed data, and the table of customers’ demographics data in the database regarding brand B transactions are used in the study of the CRM applications by constructing the following models: customers’ contribution model, customers’ segmentation model, customers’ value movement model, and customers’ churn model. To construct these models, the Recency Frequency Monetary (RFM) analysis and the Clustering method, Decision Tree, and Artificial Neural Networks data mining techniques are employed. Once these models are constructed, we then can use them to propose agile marketing strategies according to different customers’ characteristics and consuming behaviors. Keyword: Underwear Industry, CRM Applications, Data Mining, Marketing Strategies INTRODUCTION Prahalad & Ramaswamy (2000) consider the most competitive advantage does not come from the sales force or productivity but from the customer relationship. To build up the capability of managing the good customer relationship can help enterprises to understand their customers’ real need, provide their customers better service, and create the customers’ value. Therefore, enterprises do not need to do the meaningless and miserable price competition and marketing campaigns and can still win the customers’ loyalty because of the intimate customer relationship. The customers’ loyalty contributes to the enterprises profit tremendously. According to the statistics by Reichheld (1996), if the customers’ retaining rate increases 5%, their contribution to the profit increases over 25%. In addition, the cost to acquire a new customer is five times more than the cost to retain an old customer (Payne, 2002). Peppers & Rogers (1993) proposed the four steps to build up the capability of managing the good customer relationship. The first step is to identify the customers and their need and collect their information as much as possible into the customers’
  2. 2. database for later analysis. The second step is to classify the customers. By using the data mining techniques, OLAP method, and statistical analyses, the different customers’ classes can be identified and we can know what customers are the most valuable customers. The third step is the customers’ interaction. The purpose is to create interactive channels that can reach more valuable customers. Utilizing the IT technologies is a good idea to make the interactive channels more efficient. At last the fourth step is to customize the products and service. This is what the customer relationship management (CRM) aims for. By customizing the products and service to the different customers’ need, the customer relationship can be enhanced and hence the market share will be increased. Porter (1980) thought when the industry is mature the sales growth becomes smooth, the technologies are well ready, and their customers have better knowledge. At that time the mature industry should pursue the competition advantage toward the lower production cost and better customers’ service. For Taiwan’s underwear industry it is a mature industry. Every year, the total sales amount in Taiwan is about NT$ 20 billions. The half of the market is shared by the brand names from the top four companies in Taiwan’s underwear industry. However, the other half of the market is shared by no brand underwear. Therefore each company in top four is interesting in introducing CRM for better customers’ service to gain the competition advantage over the other three companies to increase its market share. There are CRM applications implemented by using data mining techniques in various industries. For example, Au & Chan (2003) mine fuzzy association rules in a bank- account database to reveal different characteristics about different customers so that they could be better served and retained. Sung & Sang (1998) proposed an application of data mining tools to hotel data mart on the Internet for database marketing. In this paper, we implement the CRM applications by constructing the customers’ contribution model, customers’ segmentation model, customers’ value movement model, and customers’ churn model for a company C in Taiwan’s underwear industry that manufactures and sells the female underwear as its major business. Among many brands owned by company C, we only study for the CRM applications on one brand named B. Therefore the database regarding brand B transactions is investigated to construct the CRM applications models by data mining techniques. To comfortably accept the models can be constructed efficiently and effectively we review the process of the data mining techniques in next section. CRISP-DM There are a lot of tools for data mining to implement CRM applications. The tool we use is the most popular one, SPSS Clementine. SPSS Clementine adopts CRISP-DM to implement CRM applications. CRISP-DM is the abbreviation of “Cross Industry Standard Process for Data Mining.” Figure 1 shows how the CRISP-DM works (The CRISP-DM Consortium, 2000). From figure 1 we can see that the CRISP-DM provides an overview of the life cycle to implement a data mining application. The life cycle consists of six phases. Below follows a brief outline of these phases. Business Understanding
  3. 3. This initial phase is to define the data mining problem by understanding the objectives and requirements of an application from a business perspective. FIGURE 1 The Process of CRISP-DM Data Understanding With the defined data mining problem in mind, the data understanding phase starts to collect the correlated data and proceeds with activities in order to get familiar with the data, to identify data quality problems, to discover first insights into the data, or to detect interesting subsets to form hypotheses for hidden information. Data Preparation The final dataset in which the data will be fed into the modeling tools is prepared in this phase. It covers all activities include the selection of tables, records, and attributes as well as the transformation and cleaning of data for modeling tools. Modeling For the same defined data mining problem type several modeling techniques are selected and applied and their parameters are adjusted to optimal values. Evaluation At this phase a model that appears to have high quality from a data analysis perspective should be determined before proceeding to final deployment for the application. To obtain the determination, it is important to more thoroughly evaluate the model and review the phases implemented to construct the model with the confidence that it properly achieves the business objectives. Deployment After the best model is determined, the deployment actions in this phase will need to be carried out in order to actually make use of the implemented model for the application.
  4. 4. The implementation order of these phases is not rigid. Implementing between different phases back and forth is always necessary depending on the outcome of each phase to the phase that has to be implemented next. CRM APPLICATIONS IN AN UNDERWEAR MANUFACTURING AND SELLING COMPAMNY We implement the CRM applications for the underwear manufacturing and selling company C by constructing the customers’ contribution model, customers’ segmentation model, customers’ value movement model, and customers’ churn model with the database regarding brand B transactions and SPSS Clementine 7.0. Brand B is designed for the female age between 40 and 60 and is a more expensive brand relative to the other brands aim at this same market segment by other companies. Its only selling channel is through department stores. Data Description Three tables in the database regarding brand B transactions are used in this study. They are the table of transactions data, the table of transactions detailed data, and the table of customers’ demographics data. In the table of transactions data we have the fields of Store ID Number, Transaction ID Number, Transaction Date, and Customer ID Number. In the table of transactions detailed data we have the fields of Store ID Number, Transaction ID Number, Transaction Date, and Expenditures. In the table of customers’ demographics data we have the fields of Customer ID Number, Zip Code, Living Area, Age, and First Time Purchasing Date. The data in this database we study have 36156 transactions by 22071 different customers purchasing brand B merchandise between January 1st, 2002 and June 5th, 2005. Customers’ Contribution Model and Segmentation Model The fields of Customer ID Number, Transaction Date, and Expenditures are used for Recency Frequency Monetary (RFM) analysis to construct the Customers’ Contribution Model. We first calculate each customer’s Recency by using the last transaction date in the database which is June 5th, 2005 to minus each customer’s recent purchasing date. Frequency is calculated by counting the times a customer purchasing brand B merchandise in the period between January 1st, 2002 and June 5th, 2005. At last Monetary is calculated by accumulating the expenditures a customer purchasing brand B merchandise in the period between January 1st, 2002 and June 5th, 2005. Once the three new fields of RFM are generated in each customer’s record, the Customers’ Contribution Model is then constructed by using the Clustering method to these fields (Kahan, 1998). After clustering the customers with these RFM fields, we can classify customers into different contribution classes. Combining this information and customers’ demographics data we construct the Customers’ Segmentation Model. In fact, two Customers’ Segmentation Models are constructed by the Decision Tree method and the Artificial Neural Networks respectively. We then choose the better performance model as our Customers’ Segmentation Prediction Model. Customers’ Value Movement Model and Churn Model In the workflow of figure 2 to construct the Customers’ Value Movement Model, we first separate the customers’ transactions data in the database we study into two sets.
  5. 5. Set 1 includes the customers’ transactions data with the transactions dates between January 1st, 2002 and December 31st, 2003 and Set 2 includes the customers’ transactions data with the transactions dates between January 1st, 2004 and June 5th, 2005. For each set we accumulate the expenditures each time a customer purchasing brand B merchandise in the period to calculate each customer’s total expenditures. The number of total expenditures is used to generate the new field of the customer’s value for each customer’s record in each data set. For example, in each data set if a customer with the number of total expenditures larger than 99% of the customers’ numbers, the field of the customer’s value in his record will be marked TOP. For a customer with the number of total expenditures larger than 95% of the customers’ numbers but smaller than 1% of the customers’ numbers, the field of the customer’s value in his record will be marked BIG. For a customer with the number of total expenditures larger than 80% of the customers’ numbers but smaller than 5% of the customers’ numbers, the field of the customer’s value in his record will be marked MEDIUM. The remaining customers’ numbers of total expenditures are smaller than 80% of the customers’ numbers and hence the field of the customer’s value in their records will be marked SMALL. The field of the customer’s value in Set 1 is named Passed Customer’s Value and the field of the customer’s value in Set 2 is named Present Customer’s Value. Therefore, if a customer with the field of Passed Customer’s Value MEDIUM and the field of Present Customer’s Value SMALL, we know this customer’s value moving downward. On the contrary, if a customer with the field of Passed Customer’s Value MEDIUM and the field of Present Customer’s Value BIG, we know this customer’s value moving upward. By combining these two fields with the fields in the table of customers’ demographics data we construct the Customers’ Value Movement Model. In fact, two Customers’ Value Movement Models are constructed by the Decision Tree method and the Artificial Neural Networks respectively. We then choose the better performance model as our Customers’ Value Movement Prediction Model. FIGURE 2 The Workflow to Construct the Customers’ Value Movement Model To construct the Customers’ Churn Model, we modify the way we mark the field of the customer’s value. For example, in each data set if a customer with the number of
  6. 6. total expenditures larger than zero, the field of the customer’s value in his record will be marked YES, otherwise, the field of the customer’s value in his record will be marked NO. Only those customers who have the field of Passed Customer’s Value marked YES are included in construction. By combining their demographics data and their associated field of Present Customer’s Value we construct the Customers’ Churn Model. In fact, two Customers’ Churn Models are constructed by the Decision Tree method and the Artificial Neural Networks respectively. We then choose the better performance model as our Customers’ Churn Prediction Model. DISCUSSION From the Customers’ Contribution Model we can distinguish customers into several groups. By studying the customers’ characteristics and consuming behaviors in each group, the different marketing strategies and interactive techniques can be applied to different groups of customers to increase the customers’ value and loyalty. As for the Customers’ Segmentation Model it can help to identify a new customer which group he belongs. Therefore in the very early stage we can develop a very unique interactive technique with him to enhance the customer relationship quickly. With the Customers’ Value Movement Model and Churn Model we can investigate why a customer’s value moves downward or just disappears and what marketing campaign is successful to what customers. In addition combining with the Customers’ Churn Model and the Customers’ Contribution Model can determine which customer has the potential customer’s value to be retained. REFERENCES Au, W. H. & Chan, C. C. 2003. Mining fuzzy association rules in a bank-account database. IEEE Transactions on Fuzzy Systems, 11, 238-248. The CRISP-DM Consortium. 2000. CRISP-DM 1.0: step-by-step data mining guide. Chicago: SPSS. Kahan, H. 1998. Using database marketing techniques to enhance your one-to-one marketing initiatives. Journal of Consumer Marketing, 15(5), 491-493. Payne, A. 2002. The value creation process in customer relationship management. Insight Interactive, 1-17. Peppers, D. & Rogers, M. 1993. Building relationships one customer at a time. New York: The One to One Future. Porter, M. E. 1980. Competitive strategy: techniques for analyzing industrial and competitions. New York: Free Press. Prahalad, C. K. & Ramaswamy, V. 200. Co-opting customer competence. Harvard Business Review, January-February, 79-87. Reichheld, F. F. 1996. The loyalty effect. Boston: Harvard Business School Press. Sung, H. H. & Sang, C. P. 1998. Application of data mining tools to hotel data mart on the Internet for database marketing. Expert Systems With Applications, 15, 1-31.

×