CHAPTER FIVE

Customer Segmentation

AN INTRODUCTION TO CUSTOMER SEGMENTATION

Customer segmentation is the process of dividing customers into distinct, meaningful, and homogeneous subgroups based on various attributes and characteristics. It is used as a differentiation marketing tool. It enables organizations to understand their customers and build differentiated strategies, tailored to their characteristics.

Traditionally organizations, regardless of the industry they operate in, tend to use market segmentation schemes that are based on demographics and value information. Over the past few decades organizations have been deciding their marketing activities and developing their new products and services based on these simple, business rule segments. In today’s competitive markets, this approach is not sufficient. On the contrary, organizations need to have a complete view of their customers in order to gain a competitive advantage. They also need to focus on their customers’ needs, wants, attitudes, behaviors, preferences, and perceptions, and to analyze relevant data to identify the underlying segments. The identification of groups with unique characteristics will enable the organization to manage and target them more effectively with, among other things, customized product offerings and promotions.

There are various segmentation types according to the segmentation criteria used. Specifically, customers can be segmented according to their value, socio-demographic and life-stage information, and their behavioral, need/attitudinal, and loyalty characteristics. The type of segmentation used depends on the specific business objective. In the following sections we will briefly introduce all the segmentation types before focusing on the behavioral and value-based segmentations that are the main topics of this book.

In general, the application of a cluster model is required to reveal the segments, particularly if we need to combine a large number of segmentation attributes. As opposed to business rules, a cluster model is able to manage a large number of attributes and reveal data-driven segments which are not known in advance.

Segments are created by analyzing the customer information residing in the organization’s data marts and data warehouse. For specific segmentation types, a market survey is required to collect the necessary segmentation criteria since the relevant attributes are unavailable (needs/attitudes, loyalty information) or partially available or outdated (demographics) in the organization’s data systems.

A market survey is also needed in order to analyze the entire market, including customers of competitor organizations. In the framework of this book, though, we are going to focus on segmenting the existing customer base.

Consumers and business customers have inherent differences, as summarized in the next section. Due to these differences, the two markets require a different segmentation approach and the use of different segmentation criteria. In this book we are going to focus mostly on segmentation for consumer markets.

SEGMENTATION IN MARKETING

Segmentation identifies the different customer typologies, facilitating the development of differentiated marketing strategies to better serve and manage the customers.

Segmentation in marketing can be used for the following:

• Greater understanding of customers to support the identification of new marketing opportunities.
• Design and development of new products/services and product bundles tailored to each segment’s characteristics rather than the mass market.
• Design of customized product offering strategies to existing customers according to each segment’s identified wants and needs.
• Offering tailored rewards and incentives.
• Selecting the appropriate advertising and communication message and channel.
• Selecting the appropriate sales and service channel.
• Determining the brand image and the key product benefits to be communicated based on the specific characteristics of each segment.
• Differentiation in customer service according to each segment’s importance.
• More effective resource allocation according to the potential return from each segment.
• Prioritization of the marketing initiatives which aim at customer retention and development according to each segment’s importance and value.

A successful segmentation scheme:

• Addresses the business objective set by the organization.
• Identifies clearly differentiated segments with unique characteristics and needs.
• Accurately assigns each individual customer to a segment.
• Provides the opportunity for profit in the identified segments.
• Comprises“identifiable”segmentswithrecognizableprofilesandcharacteristics.
• Is actionable and applicable to business operations. The organization can design effective marketing actions for the revealed segments.

SEGMENTATION TASKS AND CRITERIA

There is no magic segmentation solution that adequately covers all business situations. Different criteria and segmentation methods are appropriate for different situations and business objectives.

Segmentation is used in strategic marketing to support multiple business tasks. The starting point should always be the particular business situation. Business objectives and success criteria should be clearly defined before actually starting the analysis. The appropriate segmentation criteria and the required analytical tools should then be selected.

Table 5.1 lists the appropriate segmentation types and the required tools and techniques per business task.

SEGMENTATION TYPES IN CONSUMER MARKETS

There are various criteria for customer segmentation that can be used to optimize consumer marketing. As mentioned above, different segmentation types are used for different business situations. The following segmentation types are most widely used:

1. Value based: In value-based segmentation customers are grouped according to their value. This is one of the most important segmentation types since it can be used to identify the most valuable customers and to track value and value changes over time. It is also used to differentiate the service delivery strategies and to optimize the allocation of resources in marketing initiatives.

2. Behavioral: This is a very efficient and useful segmentation type. It is also widely used since it presents minimal difficulties in terms of data availability. The required data include product ownership and utilization data which are usually stored and available in the organization’s databases. Customers are divided according to their identified behavioral and usage patterns. This type of segmentation is typically used to develop customized product offering strategies. Also, for new product development, and the design of loyalty schemes.

3. Propensity based: In propensity-based segmentation customers are grouped according to propensity scores, such as churn scores, cross-selling scores, and so on, which are estimated by respective classification (propensity) models. Propensity scores can also be combined with other segmentation schemes to better target marketing actions. For instance, the value-at-risk segmentation scheme is developed by combining propensities with value segments to prioritize retention actions.

4. Loyalty based: Loyalty segmentation involves the investigation of the customers’ loyalty status and the identification of loyalty-based segments such as loyals and switchers/migrators. Retention actions can then be focused on highvalue customers with a disloyal profile whereas cross-selling on prospectively loyal customers.

5. Socio-demographic and life-stage: This type reveals different customer groupings based on socio-demographic and/or life-stage information such as age, income, marital status. This type of segmentation is appropriate for promoting specific life-stage-based products as well as supporting life-stage marketing.

6. Needs/attitudinal: This segmentation type is typically based on market research data and identifies customer segments according to their needs, wants, attitudes, preferences, and perceptions pertaining to the company’s services and products. It can be used to support new product development and to determine the brand image and key product features to be communicated.

Table 5.1 Segmentation types and business tasks.

Business situation/task	Appropriate segmentation criteria	Analytical tools and techniques
New product design and development	Needs/attitudinal and behavioral	Combination of data mining and market survey/factor and cluster analysis
Design of customized product offering strategies	Behavioral	Data mining/factor and cluster analysis
Brand image and key product benefits to be communicated	Needs/attitudinal	Market surveys/factor analysis and cluster analysis
Differentiated customer service	Customer value in combination with other attributes, for example, age (tenure) of customer	Binning (grouping in tiles) of customers according to their value (e.g., low n%, medium n%, top n%) and cross-tabulating with other attributes, for example, value and oldness of customer
Resource allocation and prioritization of the marketing interventions that aim at customer development and retention	Customer value supplemented by deep understanding of what drives customer decision to buy and/or to churn	Value tiles and market survey to identify drivers of decisions to buy and/or to churn
Identifying target groups for campaigns	Propensity scores derived from relevant classification models	Data mining using classification modeling – grouping customers according to their propensity scores and their likelihood to churn and/or to buy

VALUE-BASED SEGMENTATION

Value-based segmentation is the process of dividing the customer base according to value. It should be emphasized that this is not a one-off task. It is vital for the organization to be able to track value changes across time. The organization should monitor and, if possible, intervene in order to prevent downward and encourage upward migrations.

A prerequisite for this segmentation scheme is the development of an accurate and credible procedure for determining the value of each customer, on a periodic basis, preferably at least monthly, using day-to-day inputs on revenues and costs. Value-based segmentation is developed through simple computations and does not involve the application of a data mining model. Specifically, the identification of the value segments involves sorting customers according to their value and their binning in chunks of equal size, for example, of 10% named quantiles. These quantiles are the basis for the development of value segments of the form low n%, medium n%, top n%. A list of typical value segments is as follows:

• Gold: Top 20% of customers with the highest value.
• Silver: 30% of customers with the second highest value.
• Bronze: 50% of customers with lowest value.

A detailed methodological approach for the development of value-based segmentation will be presented in “A Guide for Value-Based Segmentation”.

BEHAVIORAL SEGMENTATION

In behavioral segmentation the segments are identified with the application of appropriate clustering models on usage/behavioral data that usually reside in the organization’s data warehouse or data marts. Thus behavioral segmentation can be implemented with a high degree of confidence and relatively low cost. Attributes that can be used for behavioral segmentation include product ownership and utilization, volume/type/frequency of transactions, payment and revenue history, and so on.

Typical behavioral segments that can be found in banking include:

• Depositors: Savings products – mostly deposit transactions using the network of branches.
• Future investors: Insurance and investment products – few payment and deposit transactions.
• Consuming borrowers: Consumer lending products (credit cards and consumer loans) – moderate to many transactions using all channels.
• Frequent travelers: All kinds of products – many transactions through different channels and many international transactions.
• Shoppers: Credit cards and other products – many transactions using mostly credit cards for purchases.
• Needs borrowers: Mortgage loans and consumer loans – mostly payment transactions using the network of branches.
• Classic users: Savings products and cards – moderate transactions mostly through branches and ATMs.
• Transactioners: Payroll savings products with low balances – many transactions mostly for making small withdrawals for everyday needs.
• Inactive: Unused savings accounts – no transactions.

Typical behavioral segments that can be found in mobile telephony include:

• Roamers: Heavy users of all available services – the key differentiating factor is that they use their cell/mobile phones to make calls from abroad.
• Superstars: Heavy users of all available services and all new cellular services (Internet, MMS, 3G, etc.).
• Professional users: Heavy voice users – increased voice usage and a very high incoming community (the incoming community is the number of distinct callers that have called the specific customer).
• Classic users: Average voice and SMS usage.
• Youth – SMS users: Heavy SMS users – they prefer using SMS to voice.
• Oldies – basic users: Voice usage only – very low incoming community.
• Inactive: No outgoing usage for a significant time period.

A detailed methodological approach for behavioral segmentation is presented in “A Guide for Behavioral Segmentation”.

PROPENSITY-BASED SEGMENTATION

Propensity-based segmentation utilizes the results of classification models such as churn or cross- and up-selling models. This type of segmentation involves simple computations and the binning of customers in groups according to their propensity scores. For instance, customers can be divided into groups of low, medium, and high churn likelihood as a result of a churn model. Churn models estimate and assign churn propensity scores to all customers. These scores denote the likelihood of defection and enable marketers to rank customers according to their churn risk. When the churn prediction model has been developed, customers are sorted according to their churn scores. Appropriate threshold values are determined and customers are assigned to low-, medium-, or high-risk groups according to their estimated churn likelihood.

Analysts can combine multiple propensity models and scores to create compound segmentation schemes. This procedure typically requires the application of a clustering model. Once the propensity scores are estimated by the relevant models, they can be used as inputs fields in the clustering model. As an example let us consider the case of multiple propensity models that estimate the likely uptake of the following banking products:

• Savings products
• Investment products
• Credit cards
• Consumer loan
• Mortgage loan.

Cluster analysis can be applied to the respective propensity scores in order to reveal the segments of customers with similar future needs for specific product bundles. This approach may provide valuable support in the development of an optimized product bundling and cross-selling strategy.

Propensity scores and respective segmentations can also be combined with other standard segmentation schemes such as value-based segments. For instance, when value segments are cross-examined with churn probability segments we have value-at-risk segmentation, a compound segmentation which can help in prioritizing the churn prevention campaign according to each customer’s value and risk of defection. An example of this segmentation is shown in Figure 5.1. Nine compound segments are created after combining the three low-, medium-, and high-value segments with the three low-, medium-, and high-risk segments.

The first segmentis the most critical one since it contains high-value customers at risk of defection. For this segment retention is the first priority.

LOYALTY SEGMENTATION

Loyalty segmentation is used to identify different groupings of customers according to their loyalty status and to separate loyals from migrators/switchers. The segments are created by the application of simple business rules and/or cluster models on survey or database information. By examining the loyalty segments an organization can gain insight into its strengths and weaknesses. A more elaborate loyalty segmentation is depicted in Figure 5.2 where loyals and migrators are further segmented according to the main reason for loyalty or migration.

Loyaltysegmentscanbeassociatedwithspecificusagebehaviorsandcustomer database attributes. To achieve this, an organization can start with a market survey to reveal the loyalty segments and then build a classification model with the loyalty segments’ field as the target. That way, it will be able to identify the behaviors associated with each loyalty segment and use the relevant classification rules to extrapolate the loyalty segmentation results to the entire customer base.

For example, the behaviors identified as related to the customer loyalty segments of a mobile telephony company can be:

• Dissatisfied defectors: Complaints made to call center.
• Active switchers: Always with the optimum program/product, frequent changes of programs/products in search of optimization, new customers.
• Lifestyle adapters: Usage pattern changes that could indicate life-stage changes.
• Inertial loyals: Always with the same program/product, no changes at all even if the program/product that they use does not fit their usage profile.
• Deliberative loyals: Always with the optimum program/product, normal changes of programs/products in search of optimization, old customers.
• Emotive loyals: Loyal but not included with either the inertials or deliberatives.

SOCIO-DEMOGRAPHIC AND LIFE-STAGE SEGMENTATION

In demographic segmentation customers are grouped according to their demographics. It is a widely used segmentation since demographics are considered to have a strong influence on the customers’ needs, preferences, and consuming/usage behaviors. We should point out, though, that in many cases people in the same demographic group may have different wants and needs as customers. The segments are typically created with simple business rules on the following customer attributes:

• Gender
• Race
• Social status
• Education
• Age
• Marital status
• Home ownership
• Family income
• Number of children/family size
• Age of children
• Occupation.

The family lifecycle segmentation is mainly determined by customers’ age, marital status, and number/age of children. Typical family lifecycle segments include:

• Young, singles
• Young families/no children
• Young families/children under 5
• Growing families/children above 5
• Families with children
• Older retired persons.

Quite often, the customers’ demographic data available in the organization’s databases are limited, and/or outdated or of questionable validity. Organizations can augment their demographic data repository by purchasing data from an external supplier, provided of course that this is permitted by the relevant legislation. Alternatively, the data can be collected by a market survey.

In the data mining framework, demographic segmentation is mainly used to enhance insight into the revealed behavioral, value-, and propensity-based segments.

The life-stages are defined by special events that determine a person’s priorities and main concerns, such as the birth of their first child, a significant increase in income, and so on. Life-stages present opportunities for promoting products and services that address the particular needs of customers.

An organization should try to identify the important life-stage events and link them to consuming behaviors. For example, in banking, the birth of a child is strongly related to consumer and/or mortgage loan purchases. A significant increase in annual income almost always creates investment and insurance needs and opportunities.

Since quite often the fields required for life-stage segmentation are not available in the organization’s customer databases, analysts usually try to link the age of customers to specific life-stage events.

An example from the banking industry is shown in Figure 5.3.

NEEDS/ATTITUDINAL-BASED SEGMENTATION

Needs/attitudinal-based segmentation is used to investigate customers’ needs, wants, attitudes, perceptions, and preferences. Relevant information can only be collected through market surveys and the segments are typically identified by the application of a cluster model on gathered questionnaire responses.

In the data mining framework, needs/attitudinal segmentation is mainly used in combination with behavioral and value-based segments to enrich the profile of the revealed segments, provide insight into their “qualitative” characteristics, and hence support:

• New product design/development
• The communication of the product features/brand image important for each segment
• Tailored communication messages
• New promotions.

SEGMENTATION IN BUSINESS MARKETS

Consumers and business customers have inherent differences. A business market has fewer customers and larger transactions. A decision maker, who is typically not the actual user, makes decisions for a large number of users. Selling is a long and complex process and involves talks and negotiations with people who are not the end users. Furthermore, quite often there is little association between satisfaction level and customer loyalty.

In business markets almost every customer needs a customized product, quantity, or price. In fact, each customer can be considered as a distinct segment. The segmented marketing that successfully works in consumer markets is not effective in business markets where a one-to-one marketing approach is appropriate.

These fundamental differences impose a different segmentation approach for the business markets and the use of different segmentation criteria. For instance, segmenting customers by their behavioral and usage characteristics and communicating the product features important for each segment have limited effectiveness with business customers.

The segmentation criteria typically used for business customers include:

• Value (revenue or profit)
• Size (number of employees, number of subscriptions/subscribed employees, etc.)
• Industry (government, education, telecommunications, financial, etc.)
• Business life-stage (new business, mature business, etc.).

More specifically:

1. Value based: Customer value is one of the most important segmentation criteria for both consumer and business markets. In business markets, “customer” value is typically measured at multiple levels:

(a) Per employee/subscription level for all subscribed employees within one corporate customer.

(b) Per corporate (corporate account) level.

(c) Average profit of all subscriptions/employees within one corporate customer.

(d) Per industry.

Consequently, value segmentation is also tackled at multiple levels. Usually, the differences in terms of value among corporate customers are quite large. The need to build and establish loyalty among the top-value customers is critical since, in general, they account for a substantial part of the organization’s total profit.

2. Size based: Business customers are usually ranked and grouped according to their size, based on either their total number of employees or their total number of subscriptions (subscribed employees). Table 5.2 describes a relevant size-based segmentation scheme.

3. Industry based: Corporate customers can also be divided according to their industry with the use of a standard industrial classification system. A relevant high-level industrial categorization is listed in Table 5.3.

4. Business life-stage: Corporate customers can also be segmented according to their business life-stage, and the design of the marketing strategy should be differentiated accordingly. Typical business life-stages include the startup, growth, and maturity phases (Figure 5.4).

Table 5.2 Size-based segments in business markets.

Size category	Size-based segment	Definition
Large	• Multinational accounts	• Multinational
	• Government accounts	• Government
	• Corporate	• Number of employees greater than 200
Medium	• Business	• Number of employees between 50 and 200
	• SME (Small and Medium Enterprise)	• Number of employees between 10 and 50
Small	• SOHO (Small Office, Home Office)	• Number of employees up to 10
	• Self-employed	• Independent professionals – freelancers

Table 5.3 Industry-based segments in business markets.

Technology offensive industries:

IT/telecom

Bank/finance/insurance

Business services/real estate

Media

Shipping/oil/offshore

Interest organizations/others

Technology defensive industries:

Construction

Transportation

Primary industries

Manufacturing and power supply

Trade/shop/hotel/restaurant

Public sector:

Research and development

Health and social services

Public administration/police/defense

A GUIDE FOR BEHAVIORAL SEGMENTATION

A segmentation project starts with the definition of the business objectives and ends with the delivery of differentiated marketing strategies for the segments. In this section we will focus on behavioral segmentation and present a detailed methodological approach for the effective implementation of such a project.

This section also includes guidelines for the successful deployment of the segmentation results and presents ways for using the revealed segments for effective marketing.

BEHAVIORAL SEGMENTATION METHODOLOGY

The proposed methodology for behavioral segmentation includes the following main steps which are presented in detail in the following sections:

1. Business understanding and design of the segmentation process.

2. Data understanding, preparation, and enrichment.

3. Identification of the segments with cluster modeling.

4. Evaluation and profiling of the revealed segments.

5. Deployment of the segmentation solution, design, and delivery of the differentiated strategies.

The sequence of stages is not rigid. Lessons learned in each step may lead analysts to review previous steps.

Business Understanding and Design of the Segmentation Process

This phase starts with understanding the project requirements from a business perspective. It involves knowledge-sharing meetings and close collaboration between the data miners and marketers involved in the project to assess the situation, clearly define the specific business goal, and design the whole data mining procedure. In this phase, some crucial questions must be answered and decisions on some very important methodological issues should be taken. Tasks in this phase include:

1. Definition of the business objective: The business objective will determine all the next steps in the behavioral segmentation procedure. Therefore it should be clearly defined and translated to a data mining goal. This translation of a marketing objective to a data mining goal is not always simple and straightforward. However, it is one of the most critical issues, since a possible misinterpretation can result in failure of the entire data mining project.

It is very important for the analysts involved to explain from the outset to everyone involved in the data mining project the anticipated final deliverables and to make sure that the relevant outcomes cover the initially set business requirements.

2. Selection of the appropriate segmentation criteria: One of the key questions to be answered before starting the behavioral segmentation is what attributes are to be used for customer grouping. The selection of the appropriate segmentation criteria depends on the specific business issue that the segmentation model is about to address. The business needs imply, if not impose, the appropriate input fields. Usually people with domain knowledge and experience can provide suggestions on the key attributes related to the business goal of the analysis. All relevant customer attributes should be identified, selected, and included in the segmentation process. Information not directly related to the behavioral aspects of interest should be omitted.

For instance, if a mobile telephony operator wants to group its customers according to their use of services, all relevant fields, such as the number and volume/minutes of calls by call type, should be included in the analysis. On the contrary, customer information related to other aspects of customer behavior, such as payment or revenue information, should be excluded from the segmentation.

3. Determination of the segmentation population: This task involves the selection of the customer population to be segmented. An organization may decide to focus on a specific customer group instead of the entire customer base. In order to achieve more refined solutions, groups that have apparent differences, such as business or VIP customers and typical consumer customers, are usually handled by separate segmentations.

Similarly, customers belonging to obvious segments, such as inactive customers, should be set apart and filtered out from the segmentation procedure in advance. Otherwise, the large differences between active and inactive customers may dominate the solution and inhibit identification of the existing differences between active customers.

If the size of the selected population is large, a representative sample could be selected and used for model training. In that case, though, a deployment procedure should be designed, for instance through the development of a relevant classification model, which will enable the scoring of the entire customer base.

4. Determination of the segmentation level: The segmentation level defines what groupings are about to be revealed, for instance groups of customers, groups of telephone lines (MSISDNs in mobile telephony), and so on. The selection of the appropriate segmentation level depends on the subsequent marketing activities that the segments are about to support. It also determines the aggregation level of the modeling dataset that is going to be constructed.

Data Understanding, Preparation, and Enrichment

The investigation and assessment of the available data sources is followed by data acquisition, integration, and processing for the needs of segmentation modeling. The data understanding and preparation phase is probably the most time-consuming phase of the project and includes tasks such as:

1. Data source investigation: The available data sources should be evaluated in terms of accessibility and validity. This phase also includes initial data collection and exploration in order to understand the available data.

2. Defining the data to be used: The next step in the procedure involves the definition of the data to be used for the needs of the analysis.

The selected data should cover all the behavioral dimensions that will be used for the segmentation as well as all the additional customer information that will be used to gain deeper insight into the segments.

3. Data integration and aggregation: The initial raw data should be consolidated to create the final modeling dataset that will be used for identification of the segments. This task typically includes the collection, filtering, merging, and aggregation of the raw data. But first the structure of the modeling dataset should be defined, including its contents, time frame of used data, and aggregation level.

For behavioral segmentation applications, a recent “view” of the customers’ behavior should be constructed and used. This “view” should summarize the behavior of each customer by using at least six months of recent data (Figure 5.5).

The aggregation level of the modeling dataset should correspond to the required segmentation level. If the goal, for instance, is to segment bank customers, then the final dataset should be at a customer level. If the goal is to segment telephone lines (MSISDNs), the final dataset should be at a line level. To put it in a simpler way, clustering techniques reveal natural groupings of records. So no matter where we start from, the goal is the construction of a final, one-dimensional, flat table, which summarizes behaviors at the selected analysis level.

This phase concludes with the retrieval and consolidation of data from multiple data sources (ideally from the organization’s mining data mart and/or MCIF) and the construction of the modeling dataset.

4. Data validation and cleaning: A critical issue for the success of any data mining project is the validity of the used data. The data exploration and validation process includes the use of simple descriptive statistics and charts for the identification of inconsistencies, errors, missing values, and outlier (abnormal) cases. Outliers are cases that do not conform to the patterns of “normal” data. Various statistical techniques can be used in order to fill in (impute) missing or outlier values. Outlier cases in particular require extra care. Clustering algorithms are very sensitive to outliers since they tend to dominate and distort the final solution. For general purpose behavioral segmentations, the outlier cases can also be filtered out so that the effect of “noisy” records in the formation of the clusters is minimized.

Problematic values, particularly demographic information, can also be imputed or replaced by using external data, provided of course the external data are legal, reliable, and can be linked to the internal data sources (e.g., through the VAT number, post code, phone number, etc.).

5. Data transformations and enrichment: This phase deals with the enrichment of the modeling dataset with derived fields such as ratios, percentages, averages, and so on. The derived fields are typically created by the application of simple functions on the original fields. Their purpose is to better summarize customer behavior and convey the differentiating characteristics of each customer. This is a critical step that depends greatly on the expertise, experience, and “imagination” of the project team since the development of an informative list of inputs can lead to richer and more refined segmentations.

The modeling data may also require transformations, specifically standardization, so that the values and the variations of the different fields are comparable. Clustering techniques are sensitive to possible differences in the measurement scale of the fields. If we do not deal with these differences, the segmentation solution will be dominated by the fields measured in larger values. Fortunately, many clustering algorithms offer integrated standardization methods to adjust for differences in measurement scales. Similarly, the application of a data reduction technique like principal components analysis (PCA) or factor analysis also provides a solution, since the generated components or factors have standardized values.

6. Data reduction using PCA or factor analysis: The data preparation stage is typically concluded by the application of an unsupervised data reduction technique such as PCA or factor analysis. These techniques reduce the data dimensionality by effectively replacing a typically large number of original inputs with a relatively small number of compound scores, called factors or principal components. They identify the underlying data dimensions by which the customers will be segmented. The derived scores are then used as inputs in the clustering model that follows. The advantages of using a data reduction technique as a data preprocessing step include:

(a) Simplicity and conceptual clarity. The derived scores are relatively few, interpreted, and labeled. They can be used for cluster profiling to provide the first insight into the segments.

(b) Standardization of the clustering inputs, a feature that is important in yielding an unbiased solution.

(c) Equal contributions from the data dimensions to the formation of the segments.

Identification of the Segments with Cluster Modeling

Customers are divided into distinct segments by using cluster analysis. The clustering fields, typically the component scores, are fed as inputs into a cluster model which assesses the similarities between the records/customersand suggests a way of grouping them. Data miners should try a test approach and explore different combinations of inputs, different models, and model settings before selecting the final segmentation scheme.

Different clustering models will most likely produce different segments and this should not come as a surprise. Expecting a unique and definitive solution is a sure recipe for disappointment. Usually the results of different algorithms are not identical but similar. They seem to converge to some common segments. Analysts should evaluate the agreement level of the different models and examine which aspects disagree. In general, a high agreement level between many different cluster models is a good sign for the existence of discernible groupings.

The modeling results should be evaluated before selecting the segmentation scheme to be deployed. This takes us to the next stage of the behavioral segmentation procedure.

Evaluation and Profiling of the Revealed Segments

In this phase the modeling results are evaluated and the segmentation scheme that best addresses the needs of the organization is selected for deployment. Data miners should not blindly trust the solution suggested by one algorithm. They should explore different solutions and always seek guidance from the marketers for selecting the most effective segmentation. After all, they are the ones who will use the results for segmented marketing and their opinion on the future benefits of each solution is critical. The selected solution should provide distinct and meaningful clusters that can indicate profitable opportunities. Tasks in this phase include:

1. “Technical” evaluation of the clustering solution: The internal cohesion and separation of the clusters should be assessed with the use of descriptive statistics and specialized technical measures (standard deviations, interclass and intraclass distances, silhouette coefficient, etc.) such as the ones presented in the previous chapter. Additionally, data miners should also examine the distribution of customers in the revealed clusters as well as consistency of the results in different datasets. All these tests assess the segmentation solution in terms of “technical” adequacy. Additionally, the segments should also be assessed from a business perspective in terms of actionability and potential benefits. To facilitate this evaluation, a thorough profiling of the segments’ characteristics is needed.

2. Profiling of the revealed segments: A profiling phase is typically needed in order to fully interpret the revealed segments and gain insight into their structure and defining characteristics. Profiling supports the business evaluation of the segments as well as the subsequent development of effective marketing strategies tailored for each segment.

Segments should be profiled by using all available fields as well as external information. The description of the extracted segments typically starts with the examination of the centroids table. The centroid of each cluster is a vector defined by the means of its member cases on the clustering fields. It represents the segment’s central point, the most representative case of the segment. The profiling phase also includes the use of simple reporting and visualization techniques for investigating and comparing the structures of the segments. All fields of interest, even those not used in the formation of the segments, should be cross-examined with them to gain deeper insight into their meaning. However, analysts should always look cautiously at the demographics of the derived segments. In many cases the demographic information may have not been updated since the customer’s first registration. Analysts should also bear in mind that quite often the person using the service (credit card, mobile phone, etc.) may not be the same one registered as a customer.

3. Cluster profiling with supervised (classification) models: Classification models can augment reporting and visualization tools in the profiling of the segments. The model should be built with the segment assignment field as the target and the profiling fields of interest as inputs. Decision trees in particular, due to the intuitive format of their results, are typically used to outline the segment profiles.

4. Using marketing research information to evaluate and enrich the behavioral segments: Marketing research surveys are typically used to investigate the needs, preferences, opinions, lifestyles, perceptions, and attitudes of the customers. They are also commonly used in order to collect valid and updated demographic information. It is strongly recommended that the data mining-driven behavioral segments are combined with the market research-driven demographic and/or needs/attitudinal segments. While each approach helps with understanding certain aspects of the customers, combining the approaches provides deeper insight.

For instance, provided a behavioral data mining segmentation has been implemented, random samples can be extracted from each segment and, through surveys and/or qualitative research and focus group sessions, valuable insight can be gained concerning each segment’s needs and preferences. Alternatively, the data mining and the market research approaches can be implemented independently and then cross-examined, not only as a means for evaluating the solutions, but also in order to construct a combined and integrated segmentation scheme which would provide a complete view of the customers.

In conclusion, combining data mining and market research techniques for customer segmentation can enable refined subsequent marketing strategies, based on a thorough understanding of customer behavior and needs, as shown in Figure 5.6.

Consider customers belonging to the same behavioral segment but having diverse needs and perceptions. This information can lead to tailored marketing strategies within each behavioral segment.

5. Labeling the segments based on their identified profiles: The profiling and interpretation process ends with the labeling of the identified segments with names that appropriately designate their unique characteristics. Each segment is assigned an informative and revealing name, for instance “Business Travelers,” instead of “Segment 1.” The naming of the segments should take into account all the profiling findings. These names will be used for communicating the segments to all business users and for loading them onto the organization’s operational systems.

Deployment of the Segmentation Solution, Design, and Delivery of Differentiated Strategies

The segmentation project concludes with the deployment of the segmentation solution and its use in the development of differentiated marketing strategies and segmented marketing:

1. Building the customer scoring model for updating the segments: The deployment procedure should enable customer scoring and updating of the segments. It should be automated and scheduled to run frequently to enable the monitoring of the customer base over time and the tracking of segment migrations. Moreover, because nowadays markets change very rapidly, it is evident that a segmentation scheme can become outdated within a short time. Refreshment of such schemes should be made quite often. This is why the developed procedure should also take into account the need for possible future revisions.

2. Building a decision tree for scoring – fine tuning the segments: A decision tree model can be used as a scoring model for assigning customers to the revealed clusters. The derived classification rules are understandable and provide transparency compared to scoring with the cluster model. More importantly, business users can easily examine them and possibly modify them to fine-tune the segmentation based on their business expertise. This combined approach can create stable segments unaffected by seasonal or temporary market conditions.

3. Distribution of the segmentation information: Finally, the deployment procedure should also enable distribution of the segmentation information throughout the enterprise and its “operationalization.” Therefore it should cover uploading of the segmentation information to the organization’s databases, as well as operational systems, in order to enable customized strategies to be applied across all customers’ touch points.

TIPS AND TRICKS

The following tips should be taken into account when planning and carrying out a segmentation project:

• Take into account the core industry segments before proceeding with the segmentation project and then decide which core segments need further analysis and sub-segmentation.
• Clean your data from obvious segments (e.g., inactive customers) before proceeding with the segmentation analysis.
• Always bear in mind that eventually the resulting model will be deployed. In other words, it will be used for scoring customers and for supporting specific marketing actions. Thus, when it comes to selecting the population to be used for model training, do not forget that this is the same population that will be scored and included in a marketing activity at the end. So sometimes it is better to start with the end in mind and consider who we want to score, segment, or classify at the end: the entire customer base, consumer customers, only high-value customers, and so on. This deployment-based approach can help us to resolve ambiguities about selection of the modeling dataset population.
• Select only variables relevant to the specific business objective and the particular behavioral aspects you want to investigate. Avoid mixing all available inputs in an attempt to build a “magic” segmentation that will cover all aspects of a customer’s relationship with the organization (e.g., phone usage and payment behavior).
• Avoid using demographic variables in a behavioral segmentation project. Mixing behavioral and demographical information may result in unclear and ambiguous behavioral segments since two customers with identical demographic profiles may have completely different behaviors.
• Consider the case of a father who has activated a mobile phone line for his teenage son. In a behavioral segmentation solution, based only on behavioral data, this line would most likely be assigned to the “Young – SMS users” segment, along with other teenagers and young technophile users. Therefore we might expect some ambiguities when trying to examine the demographic profile of the segments. In fact, this hypothetical example also outlines why the use of demographic inputs should be avoided when the main objective is behavioral separation.
• ‘Smooth’ your data. Prefer to use monthly averages, percentages, ratios, and other summarizing KPIs that are based on more than one month of data.
• A general recommendation on the time frame of the behavioral data to be used is to avoid using less than 6 months and more than 12 months of data in order to avoid founding the segments on unstable/volatile or outdated behaviors.
• Try different combinations of input fields and explore different models and model settings. Build numerous solutions and pick the one that best addresses the business goal.
• Labeling the segments needs extra care. Bear in mind that this label will characterize the segments, so a hasty naming will misguide all recipients/users of this information. A plain label will unavoidably be a kind of derogation, as it is impossible to incorporate all the differentiating characteristics of a segment. Yet, a carefully chosen name may simply and successfully communicate the unique characteristics of the segments to all subsequent users.
• Always prefer supervised to unsupervised models when your business problem concerns the classification of customers into categories known in advance. Predicting events (e.g., purchase of a product, churn, defaulting) is a task that can be addressed much more efficiently by classification models. A decision tree that estimates the event’s propensities will produce better results than any unsupervised model.
• Customers that do not fit well to their segment should be set apart and assigned to an “unclassified” segment in order to improve the homogeneity and quality of the segments. For instance, a customer with a very low amount in stocks and a stocks-only product portfolio, even if classified as an “Investor” by a clustering algorithm, should be identified and assigned either to the “Unclassified” or to the “Passive/Inactive” group.

SEGMENTATION MANAGEMENT STRATEGY

The following roadmap (Figure 5.7) outlines the main steps that a segmentation-oriented enterprise should follow to develop an effective marketing strategy based on behavioral segments:

Initially the segmentation results should be deployed in the customer database. Customers should be scored and assigned into segments.

In this phase the involved team should:

• Fine tune the scoring procedure, assigning customers to segments.
• Create size and profiling reports for each segment.

This step involves the evaluation and positioning of the segments in the market. The possible opportunities and threats should be investigated in order to select the key segments to be targeted with marketing activities tailored to their profiles and needs.

In this phase the involved team should:

• Conduct a market survey to enrich segments with qualitative information such as:

– Needs and wants

– Lifestyle and social status.

• Analyze segments and:

– Identify competitive advantages.

– Identify opportunities and threats.

– Set marketing and financial goals.

– Select the appropriate marketing mix strategies: product, promotion, pricing, distribution channels.

• Select and define the appropriate KPIs for monitoring the performance of the segments

Effective segmentation management requires insight into the factors that drive customer behaviors. Behavioral changes, such as customer development/additional purchases and attrition, should be analyzed through data mining and market research. The important drivers for each segment should be recognized and the appropriate marketing actions/interventions that can have a positive impact on the customer relationship with the organization should be identified. Finally, a cost–benefit analysis can enable the prioritization of the marketing actions to be implemented.

The analysis per segment should include:

• Identification of the main factors of profitability
• Identification of the main cost factors
• Assessment of the importance of behavioral drivers.

In order to make the most of segmentation, specialization is needed. Thus, it is suggested that each segment is addressed by a different segment management team. Each team should build and deliver specialized marketing strategies to improve customer handling, develop the relationship with customers, and increase the profitability of each segment. The responsibilities of the team should also include the tracking of each segment with appropriate KPIs and the monitoring of competitors’ activities for each segment.

Segment management teams should take into account the different demands of each segment and work separately as well as together to design strategic marketing plans focusing on the following:

• Channel management: Offer the appropriate channel mix in order to cover each segment’s particular demands.
• Marketing services management: Optimize the effectiveness of marketing campaigns.
• Product management: Manage existing product offerings and new product development.
• Brand management: Understand customer perception of the brand, and identify and communicate the appropriate brand image and key product benefits.

A GUIDE FOR VALUE-BASED SEGMENTATION

In contrast to behavioral segmentation, which is multi-attribute since it typically involves the examination of multiple segmentation dimensions, value-based segmentation is one dimensional as customers are grouped on the basis of a single measure, the customer value. The most difficult task in such a project is the computation of a valid value measure for dividing customers, rather than the segmentation itself.

VALUE-BASED SEGMENTATION METHODOLOGY

As with any other data mining project, the implementation of value segmentation should be guided by a specific methodology, a core process model, in order to achieve an effective outcome. Inevitably the core process model for value-based segmentation has many similarities to the one described in the previous sections for the needs of behavioral segmentation. In this section we will outline the main methodological steps of the value segmentation, emphasizing the special tasks related to such projects.

Business Understanding and Design of the Segmentation Process

This phase includes the setting of the business goal, an assessment of the situation, and the design of the process. Those analysts and marketers engaged in the project should co-operate with finance department officers to jointly determine the appropriate value measure that will be used to divide the customer base. This co-operation typically involves the identification and explanation/understanding of the cost and revenue components and a preliminary assessment of their availability. It should finally conclude with agreement on a formula that will be used for calculating the value measure.

More sophisticated measures are more effective, but harder to construct and deploy, and have a greater risk of failure. Potential value measures include:

1. ARPU (Average Revenue Per Unit): A revenue-only measure which does not take into account the different product margins, the costs to manage, and the expected customer lifetime (length of relationship with the organization).

2. MARPU (Marginal Average Revenue Per Unit): Revenue minus the cost of providing the products and services used.

3. Customer profitability: Cost to manage is also taken into account.

4. Potential value: The expected profit a customer will generate through new purchases.

5. LTV (Lifetime Value): Estimation of customer’s future value which also takes into account the customer’s expected lifetime and potential value from new purchases (LTV = customer present value + potential value).

As an example, let us consider Table 5.4, which presents a list of factors, costs, and revenues that should contribute to the development of a MARPU index in mobile telephony.

Table 5.4 Profitability factors in mobile telephony.

Factor	Description	MARPU contribution
Monthly access	Monthly access fees	+
Other one-off fees	Fees that a subscriber pays once (e.g., change of SIM card)	+
GSM revenue	Revenue from GSM usage	+
GPRS revenue	Revenue from GPRS usage	+
Other VAS revenue	Revenue from other VAS	+
Incoming revenue	Revenue from incoming calls	+
Volume discount	Discount given to the subscriber	−
Customer discount	Discount given to the subscriber	−
Interconnection costs	Cost from interconnection costs between operators	−
Termination costs	Cost from termination costs to same network	−
Roaming costs	Cost from roaming calls	−
Services costs	Cost from services offered to subscribers	−
Other dealer costs	Money paid to other dealers	−

A similar list of factors that should be taken into account in the development of a MARPU index in banking is presented in Table 5.5.

Table 5.5 Profitability factors in banking.

Factor	Description	MARPU contribution
Margin	The difference between the interest gained and the interest expense	+
Funding credit	The amount earned by banks on the deposits employed	+
Cost of funds	Amount of interest expense paid for the funds used by banks	−
Loan loss provision	The amount reserved by banks to cover expected loan losses	−
Capital charge	The capital allocated multiplied by rate of return	−
Overhead	The fixed and variable expenses associated with the product (e.g., transactional)	−

The tasks of this phase also include selection of the segmentation population and level as well as selection of the value cut-points that will be used to divide customers, for example, top 1%, high 19%, medium 40%, low 40%.

Data Understanding and Preparation – Calculation of the Value Measure

Data understanding includes a thorough investigation of the availability of the required data components and the detection of possible problems that may be encountered due to missing information. After that, all relevant customer information should be consolidated, evaluated in terms of quality, cleaned, and prepared for calculating a credible value measure. This phase ends with the development of an accurate and valid formula for quantifying the customers’ value. The calculation should use input data that cover a sufficient period of time, at least six months, so that it accurately summarizes the value of each customer.

Grouping Customers According to Their Value

There is no modeling phase in value-based segmentation. Instead of modeling, this segmentation type requires simple computations for the calculation of a credible value measure and the use of this measure to divide customers accordingly.

Customers are sorted according to the calculated value measure and binned in groups of equal size, named quantiles, such as tiles of 1% (percentiles) or tiles of 10% (deciles). For instance, the partitioning in deciles would result in 10 groups of 10% each. The customers in tile 1, assuming a ranking of customers in descending order, would contain the top 10% of customers with the highest value index, whereas tile 10 would contain the 10% of customers with the lowest value.

The derived tiles may be combined to form refined segments of the form top n%, high n%, medium high n%, medium low n%, low n%, and so on. The number and size of the derived segments define the refinement level of the segmentation solution and depend on the specific needs of the organization. Customers with zero or even negative value contribution should be identified in advance, excluded from the binning procedure, and assigned to distinct groups.

Profiling and Evaluation of the Value Segments

The derived value measure should be cross-examined with other relevant customer attributes, including all the available cost and revenue figures, as a means of evaluating the validity of the calculation and the value index.

Analysts should examine the identified segments in terms of their contribution to the total profitability/revenue of the organization in order to assess their relative importance. Therefore the size, total number of customers, and total revenue/profit that each segment accounts for should be calculated and presented in tables and/or charts such as the ones shown below for a hypothetical segmentation scheme.

The contents of Table 5.6 are graphically illustrated in the profitability curve of Figure 5.8 which depicts the cumulative size and revenue of the derived segments of our hypothetical example. Quite often a substantial part of the organization’s profit/revenue stemsfrom a disproportionately small part of the customer base. The steep incline and the high “slope” of the curve on the left of the graph designate the large concentration of value on a relatively small percentage of customers.

Finally, the structure of the value segments should be investigated by examining all information of interest, including behavioral and demographic attributes. This profiling aims at identifying the defining characteristics of the value segments. It can also include a cross-examination of the value segments with other available segmentation schemes, such as behavioral, demographic, or needs/attitudinal segments. The organization should initially focus on an examination of high-value customers since their profile can provide insight that can be used to identify existing or prospective customers with high-value potential.

Deployment of the Segmentation Solution

The last step includes the design and development of a deployment procedure for updating the value segments. The procedure should be automated to allow customers’ value to be tracked over time. It should also load the generated results into the organization’s databases as well as its operational systems for subsequent use. The marketing usage of the value segments is presented next.

DESIGNING DIFFERENTIATED STRATEGIES FOR THE VALUE SEGMENTS

The value-based segments should characterize each customer throughout his or her relationship with the organization and appropriate marketing objectives should be set for each segment.

Retention should be set as a top priority for high-value customers. In many cases a relatively small percentage of high-value customers accounts for a disproportionately large percentage of the overall company profit. Therefore maximum effort and allocation of resources should be made for preventing the defection of those customers. Additionally, the organization should try to further grow its relationship with them. Development through targeted cross-and up-selling campaigns should be set as a high priority for medium- and low-value customers with growth potential. At the same time, though, driving down the cost to serve should also be considered for those customers making trivial contributions to the organization’s profitability. The main priorities by core value segment are illustrated in Figure 5.9.

Table 5.7 provides examples of how the value segments can be used to support the design of different strategies, the prioritization of the marketing interventions, and the delivery of differentiated service level.

Monitoring the value segments is vital. It should be an ongoing process integrated into the organization’s procedures. Substantial upward or downward movements across the value “pyramid” (Figure 5.10) merit special investigation.

Segment migrations can be monitored through simple reporting techniques which compare the segment assignment of the customers over two distinct time periods (Figure 5.11), for instance at present and a few months back. This before–after type of analysis, also called cohort analysis, enables the organization to assess the evolution of its customer base.

Customers with substantial upward movement seem to strengthen their relationship with the organization. Finding their “clones” through predictive modeling and identifying customers with similar profiles can reveal other customers with growth potential.

At the other end, a decline in the pyramid suggests a relationship that is fading. This decline can be considered as a signal of a “leaving” customer. In many cases, for instance when there is not a specific event which signifies the termination of the relationship with the customer, a churn event cannot be explicitly defined through recorded data. In those cases, or as a way for being proactive, analysts can use a substantial value decline to define attrition and develop a respective churn model.

SUMMARY

Customer segmentation is the process of identifying groups that have common characteristics. The main objective of customer segmentation is to understand the customer base and gain customer insight that will enable the design and development of differentiated marketing strategies.

Clustering is a way to identify segments not known in advance and split customers into groups that are not previously defined. The identification of the segments should be followed by profiling the revealed customer groupings. Profiling is necessary for understanding and labeling the segments based on the common characteristics of the members.

The criteria used to divide customers (behavioral, demographic, value or loyalty information, needs/attitudinal data) define the segmentation type. In this chapter we briefly introduced some of the most widely used segmentation types before concentrating on behavioral and value-based segmentations which constitute the main topic of the book. For these types we proposed a detailed methodological framework and suggested ways for the use of their results in marketing. The following chapters feature a set of detailed case studies which show the use of the proposed methodology in real-world behavioral and value-based segmentation applications for various major industries.