RFM Analysis and CRM

RFM Analysis

RFM analysis is a relatively old method and first introduced by Bult and Wansbeek in 1995. RFM is a common approach for customer purchase behavior understanding. It also allows the identification of good customers by segmenting customers. According ot Bult and Wansbeek, RFM is mostly used for segmentation analysis in direct marketing; through RFM, marketer can sort out target customers from a huge list of customers for its marketing activity. This makes RFM a quite popular method especially for the retailer industry.

RFM is a method for customer segmentation based on purchase behavior. It is primarily used for improving the efficiency of marketing efforts to existing customers by understanding them quantatively. There are other techniques described here such as Clustering, Classifications, and Association Rule Mining. However, the effectiveness of RFM Analysis technique is proven in studies when RFM Analysis is applied to the marketing databases. It is “a very powerful tool that involves little more than creating segments from the three groups”:

Various studies are conducted on RFM analysis and a number of improvements and changes have been proposed:

Case  -  Clothing: Underwear Industry

Hung et. al. describes their findings about a company which produces female underwear. In this study, transactions for one of the brands, Brand B, are studied. Three tables from the related database were used. These tables contain 36156 transactions by 22071 unique customers between January 1st, 2002 and June 5th, 2005:

During the study, customers’ contribution model and customers’ segmentation model has been constructed using Recency Frequency Monetary (RFM) analysis among other types of models. Recency is calculated for each customer by simply subtracting the most recent purchase date and January 1st, 2005. Frequency and Expenditure are, similarly calculated for each customer, by summing up the total number of purchases and their expenditures. When all the RFM measures are generated for each customer, a customer contribution model is constructed by using clustering method to these fields.

By combining the clusters and customers’ demographics data, two Customers’ Segmentation Models are constructed: One with Decision Tree method and the Artificial Neural Networks. Better performing segmentation model is used for predictions.

By using the Customers’ Contribution Model, customers are distinguished into several groups. Characteristics, consuming behaviors of these groups are analyzed and different marketing strategies and interaction techniques are used for each group. Having a special strategy and interaction for each group, the customer value and loyalty have been increased. With the help of Customers’ Segmentation Model, it takes less time to place a new customer into the appropriate customer group. From the beginning, new customer is treated according to his group. The relationship with this customer has been enhanced quickly.

Case - Banking: Credit Cards

Wu et. al. describes a case study to identify influential groups among the credit card owners. Authors collect credit card customer data and spending records in 2003 from banks. The study focuses on the spending behavior of customers using 1063000 pieces of purchase data for both 2003 and 2004. During the study, Weka is used to preprocess the dataset and apply these algorithms: Classification, Clustering and Association Rule Mining.

Using Weka, the customer data of 2003 is clustered to identify the influential groups in terms of business profits and behavioral pattern. 2004 dataset was used to observe changes after a year. The clustering of 2003 dataset is realized with the RFM model. Since the recency of credit card use has little significance in customer value, frequency and monetary measures have been used. As a result of the clustering, 5 clusters have been identified according to frequency and monetary measures.

It is revealed that 2 clusters (80% of the customers) had spent less than $4500, meaning that they bring little profit to the bank. Bank should consider not paying too much energy and cost to these types of customers. 2 of the remaining clusters were loyal to the bank but did not bring high profits. Bank should focus on marketing and enhance the campaigns on these customers for a better customer profile. The last cluster, 3.3% of the customers, was the gold customers with high loyalty and high profit. Losing such customers means a great loss for the bank and they should retain these customers with the help of promotions, service improvements targeting these groups.

References

May 2013