Intelligent Value Based Customer Segmentation: A Case Study of Automobile Retailer — Summary

In this article, we will be reviewing a research paper titled: ‘Intelligent Value-Based Customer Segmentation Method for Campaign Management: A Case Study of Automobile Retailer’ (2008).  The research proposes an approach that blends consumer targeting and customer segmentation for an effective marketing strategy that targets the right kind of consumers.

The research paper is authored by Chu Chi Henry Chan. It was published by Elsevier in 2008. Chu has over 20 publications to his name. He teaches at the Chaoyang University of Technology in the Industrial Engineering and Management department.


According to the paper, there are two basic models to effectively segment customers:

1- RFM: Recency, Frequency, and Monetary Model

2- LTV: Customer Life Time Value Model

Additionally, this work uses a Generic Algorithm (GA) to further segment customers.

The automobile retailer chosen for this study was Nissan — with over four thousand customers to segment. The group was studied extensively in order to develop the most efficient marketing strategy.

The research problem that required a solution

Market segmentation can aid in the identification and comprehension of a company’s target audiences and ideal clients. If someone is a marketer, this helps to target their advertising more successfully by identifying the correct market for their goods or services.

Market segmentation can also assist marketers of different companies to increase the effectiveness of their marketing efforts by allowing businesses to reach the appropriate individuals at the right moment with the correct message. They may use segmentation to discover more about the business’ customers and personalize their content to their wants and preferences.

It is far more beneficial to target a particular group that is likely to engage in their material or service rather than a large group of people that may or may not be interested in what the company is willing to offer. If marketers advertise to the overall industry, they will spend a fortune on commercials, but only a small number of people will acquire their product.

Alternatively, if the marketers focus on a group with the correct traits, the campaign’s conversion rate may increase substantially. What we have learned is that if companies narrow down the scope of their demographics, they would be more successful in increasing their sales. 

In contrast, there is no purpose to promote toolbox kits to anyone other than the professionals and some enthusiasts. Advertising them to a massive group would be a waste of money and resources.

Companies can also employ market segmentation to provide more accurately targeted marketing alternatives and to tailor their information to specific audience segments.

The importance of successful customer market segmentation is quite evident. While this all seems very convenient for marketers, the reality is different. There is a gap between the practice of targeting customer segments and the actual campaign management.

Several previous types of studies were using a variety of mathematical methods to segment customers without taking into account the relationship between customer segmentation and campaign effectiveness. The connection between client segmentation and campaign activity was traditionally either manual or non-existent.

Marketing companies, rather than focusing on within-segment uniformity or focusing rate, are engaged with and interested in maximizing the total value of targeted clients.

Instead of assessing homogeneity inside a particular segment, this study uses a customer lifetime value (LTV) model to analyze the compatibility between intended consumer groups and marketing techniques.

The proposed solution to the research problem

As mentioned before, the fundamental issue with creating customer value is that most scientific literature is unable to connect consumer segmentation methodologies to campaign actions. This study recommends tying the relevant variables of marketing activities to consumer segmentation to incorporate these two procedures.

This inquiry examines customers using two primary factors: current value and potential value, to accomplish LTV in both phases.

Initially, it is important to decide on what kind of value one should provide to their clients.

A customer’s wants and requirements are the emphases of a target marketing plan. As a result, defining the target markets that the businesses will aim to service is a necessity for the creation of this customer-centric strategy (Camilleri, 2017).

Furthermore, it’s important to take into account which clients should receive value. Lastly, the researchers inquired about how to discover and engage appropriate consumers.

As a result, the goal of customer segmentation is to identify customers who would be good candidates for a marketing campaign. Customer segmentation should be related to the successful delivery of the intended customer value and successful campaign execution.

Customer segmentation is used in numerous marketing researches to boost customer value or revenue by carefully attracting consumers. IBM’s CRM research team offers 2W (What, Whom), and 1H (How) as three essential variables in creating customer value to attain this goal. These are summarized in Figure 1.

Figure 1: Customer Value Creation defined by IBM

For campaign tactics, this research proposed a transformational approach that blends consumer targeting and customer segmentation. This inquiry employs:

  • A recency, frequency, and monetary (RFM) model to identify customer behavior
  • Then applies a customer life time value (LTV) model to assess suggested segmented consumers
  • It is also suggested that a generic algorithm (GA) be used to choose more acceptable clients for each campaign plan

The well-known recency, frequency, and monetary (RFM) model are employed to depict customer behavior variables to determine customer behavior. 

Planners can more successfully identify important customers and establish appropriate marketing strategies by using the RFM model (Wei et al., 2010).

To characterize client behavior utilized, this method includes three attributes of customer transactional data: recency, frequency, and monetary value.

The first dimension is recency, which reflects how long it has been since a transaction began. However, Frequency is the second dimension, which reflects how commonly a buyer acquires things during a given period. Lastly, monetary value refers to the amount of money spent by a consumer during a given period.

The RFM model is based on the idea that future consumer market movements would mimic past and current trends. To highlight customer behavior patterns, the estimated RFM values are summarized. The following RFM variables are proposed for use in this study:

  • Recency (R): the amount spent most recently.
  • Frequency (F): the overall number of items purchased over a period of time.
  • Monetary (M): monetary value expended over a period of time.

This research proposed a lifetime value (LTV) model as the fitness function of GA to evaluate each cycle of customer segmentation suggested by GA. In most cases, client lifetime value must take into account both previous financial input and predicted future success.

Customers’ lifetime value should be composed of two major elements, according to this research: potential value and current value.

The total of these two numbers could be used as an indicator to measure the value of a customer’s lifecycle.

LTV models could not connect campaign-related activities. The fundamental issue with most existing studies is that the LTV model does not take into account the relationship between customer values and campaign activity.

This paper presents a novel approach for assessing customer value depending on a serial number of campaigned initiatives to address this issue.

GA (Generic Algorithm) is a stochastic method that is devoid of derivatives and is derived from natural selection and evolution. Below are given the properties of GA (Wei et al, 2010):

  • GA can be used to solve problems in both continuous and discrete optimization.
  • Because GA is stochastic, it is unlikely to become stuck in local minima.
  • In the complex model, GA’s adaptability helped with both structure and recognition.

This research involves converting input information into a binary bit string termed a chromosome to use GA in consumer segmentation. Instead of a single point, GA normally considers a group of locations as a population and assesses fitness values frequently to arrive at better results.

The encoding technique, selection, crossover, mutation, and decoding demonstrated in Figure 2 (Fig. 7 of the research article) make up GA.

Figure 2: Producing a better generation by GA.

Customers were segmented using simple GA in this research. The system flow is depicted in Figure 3 (Figure 8 in the article). The input settings must first be configured, and the customer data must be converted as a binary string.

Figure 3: System flowchart of GA.

Fourth, members with a higher fitness value are chosen as parents for the future generation.

Fifth, the crossover is employed to make new chromosomes with a demonstrable crossover rate so that beneficial genes from parents can be preserved.

Sixth, the mutation is utilized to switch a little with a fixed mutation rate probability. To avoid the overall community from collapsing on trapped local optima, this stage can manufacture new chromosomes.

On the seventh day, a new generation is born.

However, the eighth phase involves assessing a new generation to determine the stop criterion.

The main findings reported by the authors

RFM and LTV models were officially identified as two separate forms of segmenting clients in most investigations. RFM models usually describe the dynamic behavior of customers.

On the contrary, LTV models are used to evaluate the profitability or contribution of a client. In customer segmentation research, understanding the behavior of high-value consumers is a critical task.

Through the use of an LTV assessment methodology, this research offers an efficient approach that employs GA to select consumer RFM behavior percent. The corresponding value of GA is determined by the customer lifetime value. High-value clients can be recognized for campaign initiatives using the methodological approach that has been proposed.

Another advantage of the suggested approach put forward by the author of this research article is that it takes into account the relationship between customer values and marketing initiatives. As a result, valuable customers might be selected for a marketing effort.

Suggestions for future research on Customer segmentation

This step is critical for answering different questions which could further expand the research study. Furthermore, the restrictions are adaptable and applicable, meaning that they are something that other academics believe can be achieved under different conditions. This was also crucial in formulating suggestions for future studies.

Moreover, the author explained why the aforementioned topics were not explored in the present study. There’s also a list of potential roadblocks that other academics might run into when conducting further research on this subject.

This will assist them in developing a realistic or realistic strategy for their investigation, such as the scope, goal, and outcomes.

Eventually, the researcher’s method becomes more straightforward. To be more explicit, several direct study proposals for future studies were made to other experts. They are accurate enough for the reader to feel comfortable conducting further research in the indicated topics, such as:

  • The authors would like to create an investigation to compare the benefits and drawbacks of current segmentation techniques.
  • This study excludes customer acquisition because it only analyses segmenting current customers rather than acquiring new ones.
  • Future research will look into the prospect of locating and attracting new clients.

The implications for practitioners that plan to work on Customer segmentation

The direct consequence of an event is the conclusion or end outcome that happens when specified occurrences unfold. In this instance, practical refers to the actual outcomes of an action, while implication refers to the logical relationship between the occurrence and the conclusion.

Customers’ potential value, devotion, and retention rates could all be increased by using the proposed way to segment them. The study report includes several figures that explain and demonstrate the usefulness of the suggested model. The proposed strategy, for instance, increases the potential values of the segmented group, as shown in Fig. 9 in the research article.

The potential values of groups 1, 3, and 4 have been enhanced by more than a hundred percent. A similar impact may be seen in Figure 4 for loyalty. This outcome showed that the suggested strategy is successful for promotional efforts and can boost customer value.

Figure 4: The expected year of loyalty by using GA and random selection.

The situation that would emerge if particular circumstances were met is known as practical implication. For example, when researchers conduct behavioral trials, the accuracy of the data they gather has practical ramifications for doctors’ capacity to reliably identify the efficacy of specific behavioral therapies.

The end findings show that the proposed strategy can boost potential value, customer loyalty, and customer lifetime value. Nevertheless, there are a handful of flaws in this research:

  • To begin, the suggested solution necessitates a large amount of client information.
  •  A Nissan dealer spent around six months collecting a decade worth of client data for this investigation. As a result, implementing this strategy is a difficult task for a business.

Moreover, each variable has just one threshold. More breakpoints will be studied later to establish the best number of breakpoints for segment customers. 


  • Camilleri, M. A. (2017, 12 01). Market Segmentation, Targeting, and Positioning. Researchgate.
  • Jang, J.-S., Sun, C.-T., & Mizutani, E. (1997). Neuro-Fuzzy and Soft Computing-A Computational Approach to Learning and Machine Intelligence.
  • Wei, J.-T., Lin, S.-Y., & Wu, H.-H. (2010, 01 09). A review of the application of RFM model. Researchgate.