Journal of Industrial Engineering, University of Tehran, Special Issue, 2011, PP. 79-93 79
Analyzing Customer Segmentation Based on Customer
Value Components (Case Study: A Private Bank)
Mahboubeh Khajvand *1 and Mohammad Jafar Tarokh 1
1 IT Group, Industrial Engineering Department, K.N. Toosi University of Technology, Tehran, Iran
(Received 6 December 2010, Accepted 12 June 2011)
Studying about the customer segmentation and begetting customer ranking plan diverts more attention in recent years. In this regard, this study tries on providing a methodology for segmenting customers based on their value driver parameters which extracted from transaction data and then ranks customers with regard to their customer lifetime value (CLV) score. Discovering hidden pattern between customers ranking result the other data such as customer product ownership data and socio-demographic information is the other work which addressed in this paper. Achieving this, we used data mining techniques such as different classification and clustering approaches, and implemented them on real data from an Iranian private bank. Current study can provide good insights for marketing and CRM department of the organization in identifying different segments of customer for designing future strategic program.
Keywords: Customer lifetime value, Customer segmentation, Data mining, RFM analysis, Decision rule
Nowadays, with increase in market competition, more and more companies do realize that their most priceless asset is the existing customers , so they give more attention to customer relationship management (CRM) than past. The main goal of a CRM system is to understand profitable customers, to create and sustain relations with them. Thus it is so important to segment customers based on their value and dedicate rank to them and establish different relation with different segments of customer and different ranks. Customer lifetime value (CLV) as a paradigm of analytical CRM is considered by researchers and companies in different industries. CLV is the present value of all future profits for firms generated from a customer . Calculating CLV has had lots of applications and several authors have developed models for different applications such as performance measurement , targeting customers , marketing resources allocation [5,6], product offering [7, 8, 9], pricing , and customer segmentation [11, 12, 13].
This study prepared to segment customers of an Iranian bank and based on it, the bank can focus on niche segment to propose new product and services to them which is one of the determined decisions in marketing strategy.
* Corresponding author: Tel: +98-21-84063362 Tel: +98-21-88674858 Email: [email protected]
We considered three factors: Recency (R), Frequency (F), and Monetary (M) to cluster customers, analyzing clusters in RFM aspect, calculating CLV value of different clusters. Then clusters with homogeneous CLV value incorporate and construct a segment and based on the CLV value of the segments, we dedicate rank to them. After that decision tree and decision rules classifiers are used to estimating the accuracy rate of the customer segmentation and present rules for different segments to provide a clear and perspicuous knowledge explanation about different segments. Then socio-demographic predictors (e.g. age, sex) and product ownership predictors (e.g. types of customer accounts) are employed as input variables to extract rule set for segments. These rules explain which customer with which characteristics lead to which segment with which value. Then evaluate which of the predictors have positive, negative, or neutral effect on customer CLV rank. Based on these extracted knowledge the bank could develop branch strategies like credit endowment or facility grant, marketing strategies, CRM strategies or even organization strategies for different segment of customer with different CLV rank. The rest of this study is organized as follows. Section 2 outlines the background and reviews related work on customer lifetime value, CLV divisions and classifications, RFM analysis, and data mining definitions and techniques in customer segmentation application. Section 3 describes the research methodology and case study. Finally, section 4 draws conclusions and summarizing the contributions of this work.
1.1. Customer lifetime value
Customer lifetime value (CLV) concept is going from customer relationship management (CRM) issue. Peppers believes that “the goal of CRM is to forge closer and deeper relationships with customers and to maximize the lifetime value of a customer to an organization” . There has been an explosion of interest in the discipline and practice of CRM in the worlds of business and academic over the last decade especially in identifying and ranking of customers based on customer value drivers. Based on the approach of estimating CLV, there are different definitions for this term. One of the earliest definition said, CLV is expected profits from customers except cost of customer management . Pfeifer et al. defined CLV as the present value of the future cash flows attributed to the customer relationship  and finally Sublaban and Aranha  described CLV as estimated monetary value that the client will bring to the firm during the entire lifespan of his/her commercial relationship with the company, discounted to today value. In literature review, there are some classifications for CLV models. One of these divisions was proposed by Jain and Singh in  and the other by Gupta et al. in . Jain and Singh determined that many models have been proposed in CLV literature dealing with all kinds of issues related to CLV. The following selection of models provides summaries of some key models addressing some major research opportunities in CLV research and applications. Based on the threefold stream of research related to CLV, they divided them into three corresponding categories :
Models for calculation of CLV: This category includes models that are specifically formulated to calculate the CLV and/or extend this calculation to obtain optimal methods of resource allocation to optimize CLV.
Models of customer base analysis: Such models take into account the past purchase behavior of the entire customer base in order to come up with probabilities of purchase in the next time period.
Normative models of CLV: These models have been proposed and used mainly to understand the issues concerning CLV. Managers depend on many commonly held beliefs in making decisions regarding CLV.
Proposed paper works on normative model of Jain and Singh categories. The result of this research could be used by different department of the bank to make decision or plan strategy.
Gupta et al. described six modeling approaches in CLV issue :
RFM Models: Based on Recency,
Frequency, and Monetary.
Probability Models: Based on Pareto/NBD model and Markov chains.
Econometric Models: Like probability model based on Pareto/NBD model and customer acquisition,
Analyzing Customer Segmentation ….. 81
customer retention, and customer margin and expansion.
Persistence Models: Based on modeling the behavior of its components, that is, acquisition, retention, and crossselling.
Computer Science Models: Based on theory (e.g., utility theory) and are easy to interpret. In contrast, the vast computer science literature in data mining, machine learning, and nonparametric statistics has generated. Diffusion/Growth Models: Based on customer equity (CE).
This study works on RFM model and uses computer science models’ technique of Gupta’s categories.
1.1.1. RFM analysis
One of the most powerful and simplest models to implement CRM may be the RFM model – Recency, Frequency, and Monetary value . Bult and Wansbeek defined RFM as : (1) R (Recency): the period since the last purchase; a lower value corresponds to a higher probability of the customer´s making a repeat purchase; (2) F (Frequency): number of purchases made within a certain period; higher frequency indicates greater loyalty; (3) M (Monetary): the money spent during a certain period; a higher value indicates that the company should focus more on that customer.
In recent researches, some authors proposed WRFM – Weighted RFM – instead of RFM. Depend on the importance of these parameters in their case they dedicated weights to R, F, and M. For example, Stone in  suggested placing the highest weighting on the Frequency, followed by the Recency, with the lowest weighting on the Monetary measure, in Liu and Shih  study, Recency is the most important parameter and Monetary is the less important parameter, but in Chuang and Shen study, Monetary has the most value and Recency had the least value .
To determine importance (weight) of RFM parameters, AHP method is exploited.
The three main steps of this method are as follows (:
Perform pairwise comparisons with asking evaluators (decision makers or
Assessing the consistency of pairwise judgments;
Employing eigen value computation to derive the weights of RFM variables.
Some researches try to develop RFM model and add some parameters to these three parameters. For example, Cheng Yeh et al. derived an augmented RFM model, called RFMTC model (Recency, Frequency, Monetary value, Time since first purchase, and Churn probability), using Bernoulli sequence in probability theory . In this study, RFM parameters are employed for customer segmentation.
1.2. Data mining concept and methods Simply stated, data mining is the process of automatically discovering useful information in large data repositories. Imielinski and Virmani described data mining as a pattern query in large data bases . Data mining should have been more appropriately named knowledge mining from data. Han and Kumber believe that it is a step in the knowledge discovery process . Knowledge discovery as a process is depicted in an iterative sequence of the following steps: Data cleaning; Data integration; Data selection; Data transformation; Data mining; Pattern evaluation; Knowledge presentation. The relation between the steps is shown in figure