Ch11-H8566.qxd 8/8/07 2:04 PM Page 222

CHAPTER 11 Market segmentation

YORAM (JERRY) WIND and DAVID R. BELL

All markets are heterogeneous. This is evident from the focus of a significant part of the marketing observation and from the proliferation of popular research literature. The basic concept of segmenta- books describing the heterogeneity of local and tion (as articulated, e.g. in Frank et al., 1972) has global markets. Consider, for example, The Nine not been greatly altered. And many of the funda- Nations of North America (Garreau, 1982), Latitudes mental approaches to segmentation research are and Attitudes: An Atlas of American Tastes, Trends, still valid today, albeit implemented with greater Politics and Passions (Weiss, 1994) and Mastering volumes of data and some increased sophistica- Global Markets: Strategies for Today’s Trade Globalist tion in the modelling method. To see this, consider (Czinkota et al., 2003). When reflecting on the nature the most compelling and widely used approach to of markets, consumer behaviour and competitive product design and market segmentation – con- activities, it is obvious that no product or service joint analysis. The essence of the approach out- appeals to all consumers and even those who pur- lined in Wind (1978) is still evident in recent work chase the same product may do so for diverse rea- by Toubia et al. (2007) that uses sophisticated geo- sons. The Coca Cola Company, for example, varies metric arguments and algorithms to improve the levels of sweetness, effervescence and package size efficiency of the method. Other advances use for- according to local tastes and conditions. Effective mal economic theory to specify optimal consumer marketing and business strategy therefore requires trade-offs – see Iyengar et al. (2007) for an applica- a segmentation of the market into homogeneous tion to non-linear pricing. segments, an understanding of the needs and wants Despite the underlying stability of the basic of these segments, the design of products and serv- concept, recent advances in information technology ices that meet those needs and development of and the trend towards globalization are introducing marketing strategies, to effectively reach the target a discontinuous change to the adoption and imple- segments. Thus focusing on segments is at the mentation of segmentation strategies. The revolu- core of organizations’ efforts to become customer tion in information technology and strategy makes driven; it is also the key to effective resource alloca- possible the creation of databases on the entire uni- tion and deployment. The level of segment aggre- verse and enormous advances in database market- gation is an increasingly important issue. In today’s ing and innovative distribution approaches. It has global economy, the ability to customize products also facilitated much of the development in flexible and services often calls for the most micro of seg- manufacturing with the consequent emergence of ments: the segment of one. Following and imple- mass customization. In addition, the Internet has menting a market segmentation strategy allows the expanded not only the ability to implement market firm to increase its profitability, as suggested by the segmentation research more effectively, but also classic price discrimination model which provides expanded the portfolio of segmentation methods the theoretical rationale for segmentation. available for use (see, e.g. Dahan and Srinivasan, Since the early 1960s, segmentation has been 2000). These changes are leading to the creation of viewed as a key marketing concept and has been ‘one-on-one marketing’ or segments of one. The Ch11-H8566.qxd 8/8/07 2:04 PM Page 223

Market segmentation 223

globalization of business expands the scope of allocation and differential action based on seg- operations and requires a new approach to local, ment (customer) value. The chapter concludes regional and global segments. Moreover, busi- with a set of critical issues that provide the guide- nesses that have not traditionally embraced mar- lines for research agenda in this area. keting in general or segmentation in particular, see it as imperative for success and even survival. Consider the current enormous effort by the lead- Use of segmentation in marketing ing financial services firms to understand how to segment the global at-retirement market fuelled by and business strategy the baby boomers in North America and Western Europe. Finally, it is important to recognize a subtle Conceptually any business strategy should be but pervasive shift in the bases for segmentation. based on understanding, meeting and even Historically, marketers segment the market accord- exceeding the needs of target segments. Figure 11.1 ing to characteristics (e.g. demographics), prefer- illustrates the centrality of segmentation and the ences, usage rates, etc. Increasingly, it is difficult to progression of fundamental questions to address. fully articulate a segmentation strategy without an At the core is the identification of the existing and accompanying discussion of customer lifetime potential customer base, an understanding of value (CLV) and a thought process that makes the underlying heterogeneity and the evolving needs CLV calculation explicit (see Gupta and Lehmann, and wants of target segments. Next is the response 2003). to segmentation, namely guidelines for the devel- These changes require not only an appraisal opment of products and services, and their associ- of what we know about segmentation, and what ated positioning to meet the evolving needs of the works and does not work, but also a review of the target segments. Finally, the product positioning segmentation area as part of an entirely new mar- provides the foundation for the rest of the market- keting and management paradigm. ing strategy and the processes, resource allocation Therefore, the purpose of this chapter is to decisions and other activities of the firm. introduce the reader to both the ‘best practice’ in Numerous published and unpublished case the segmentation area and the likely new develop- studies attest to the value of segmentation. For ments. These observations are based on advances example, Bell et al. (1998) show how segmentation in marketing concepts, marketing science research of store choice decisions of supermarket shoppers and modelling tools, generalizations from empir- reveals fundamental differences in store attractive- ical studies, successful practices of leading firms ness, conditional on a shoppers preferred shopping and the conceptual implications of operating in style. The model illustrates how one store format the global information age. This discussion of the can capture market share from another. It is import- best segmentation practices and likely advances ant to recognize that applications of segmentation encompasses five areas: cover a diversity of business contexts. In an indus- trial buying setting, Gensch et al. (1990) provide 1 Use of segmentation in marketing and business compelling evidence of the positive consequences strategy. of segmentation of electrical equipment buyers. In 2 Decisions required for the implementation of a a 1-year test segmentation applied in two of three segmentation strategy. geographic districts, sales increased 18 and 12 per 3 Advances in segmentation research. cent – while sales declined 10 per cent in the district 4 Impact of operating in the global information age in which model-based segmentation was not on segmentation theory, practice and research. applied and 15 per cent for the industry. The firm 5 Expansion of segmentation to other stakeholders. reports continuous market share increases from the application of the segmentation approach. This is Thus, this chapter is based on the premise that not an isolated case. The popular business press segmentation is the firm’s response to a funda- and the conference circuit are full of anecdotal cases mental market feature – heterogeneity. The likely in which creative segmentation has paid off. In fact success (or otherwise) of the firm’s segmentation a growing number of firms do use segmentation as strategy is assessed through a segmentation audit the basis of their marketing strategy. discussed next. The firm enacts the segmentation Yet despite the general acceptability of seg- strategy through: (1) data collection, (2) applica- mentation and its value, too many firms are not tion of models and frameworks and (3) resource segmenting their markets effectively and are not Ch11-H8566.qxd 8/8/07 2:04 PM Page 224

224 The Marketing Book

1. Who are the customers and what are their needs and wants?

2. What product/service offerings will meet the target segments’ needs and offer us a sustainable competitive advantage?

3. What strategies and programmes, resources, capabilities, and processes are required to develop and effectively implement the product/service solutions?

Figure 11.1 Focus on market-driven strategy

basing their strategies on the evolving needs of tar- identity the observable characteristics of get segments. The experience of the more success- individuals or firms thought to be in the target ful firms in consumer and industrial markets alike market segments.A firm unable to effectively suggests, however, that effective segmentation is a answer this question is likely to have considerable must. The likelihood of a positive response to the trouble not only locating existing segments, but firm’s offerings is increased, the cost of reaching also predicting the evolution of new market customers and chances of new product and service segments. failures are reduced. The need to ‘rediscover’ the 2 What do my segments think and feel? An centrality of segmentation is made forcefully in a attitudinal evaluation of segments focuses on recent article by Yankelovich and Meer (2006). pyschographics and underlying segment A segmentation audit can help a firm make preferences.A firm that cannot develop an an initial determination as to whether it uses an answer to this question is unlikely to be able to effective segmentation strategy. We propose three communicate effectively with target segments.An interrelated approaches in decreasing order of com- understanding of perceptions, preferences and plexity and time commitment. Table 11.1 presents a attitudes towards the firm and its competitors fairly comprehensive audit template (based on an are necessary conditions for being able to tailor actual audit for a large computer manufacturer). In marketing messages and talk in the appropriate scoring this particular audit it is important to note ‘language’. that effective segmentation requires a positive 3 How do my segments behave? This question answer to each question. Any lower score on any of forces the firm to think about usage, demand and these dimensions requires correction. consumption patterns, and the reaction to A second approach is the ‘Five Question’ (5Q) changes in the marketing mix (product, price, method which requires the firm to articulate an promotion and distribution). answer to each of the following: 4 Where are my segments going? Here, the firm must attempt to map out the trajectory of 1 Who are my market segments? This descriptive segment growth. All segments follow a life cycle approach forces management to try and first and this question forces the firm to address Ch11-H8566.qxd 8/8/07 2:04 PM Page 225

Market segmentation 225

Table 11.1 A segmentation audit

Practice Completely Somewhat Does not Do not describes describes describe know us us us at all

1 Our business strategies recognize the need to prioritize target segments 2 Our marketing plans include specific plans for each of the selected segments 3 We have specific product offerings for each target segment 4 We have a process for updating the information on our segments on an ongoing basis 5 Our segments balance the unique country needs with potential synergies across countries 6 We have an effective process for implementing segmentation research 7 We have an effective process for implementing segmentation strategies 8 We have P&L reports and accountability by segment 9 We have detailed information about segments, including: Current size of segment Potential size of segment Key business needs of the segment Information systems needs of the segment Their prioritized needs/benefits sought Their prioritized preference for product and service features Demographic characteristics of the segments Product/system ownership and usage Competition’s strength in each segment Perceived positioning of each competitor by the members of the segment 10 Information about the target market segments are incorporated effectively into the following strategies: Positioning Product and service offerings Pricing Promotion Advertising Distribution Sales force Ch11-H8566.qxd 8/8/07 2:04 PM Page 226

226 The Marketing Book

dynamics, and thereby understand long-term Effective segmentation strategy requires detailed viability of segments. answers to the following sets of questions: 5 What are my individual customers (and segments) worth? This final question stipulates that the firm 1 How to segment the market? must attempt to place a financial value on market 2 What research procedure to use to develop a segments.The inability to answer this question is segmentation strategy? likely to hamper firm efforts to effectively allocate 3 What segment(s) to target? marketing resources and maximize marketing 4 How to allocate resources among the segments? return on investment (ROI). 5 How to implement the segmentation strategy?

A third exercise which is less time intensive, but Segment identification decision can nevertheless reveal fundamental strengths and weaknesses in the current segmentation The determination of which set of variables – basis – strategy is the ‘Product by Segment Matrix’ (PSM). to use for segmentation of the market is critical. To create the PSM, the firm simply lists as rows of Conceptually, the guiding principle is fairly obvi- a matrix, the distinct products and services cur- ous. A good segmentation variable is one that rently on offer. The columns of the matrix are a explains variation in use of the firm’s products and list of the segments thought to be addressed by services. If a proposed segmentation variable has no the firm. At the very least, this exercise will pro- correlation to choice or other important behaviours, vide some insight into the firm’s view of market it is clearly of little value. Practically, the approach is structure. For example, if all the matrix entries fall quite involved and requires consideration of the fol- on the diagonal, each product is addressing a dif- lowing issues. Should we segment on product ferent market segment. A ‘full row’ implies that usage patterns (e.g. users versus non-users or heavy one of the firm’s products is serving the needs of versus light users)? Should we segment based on multiple segments simultaneously. Such a prod- benefits sought (e.g. product performance versus uct is clearly of key strategic importance. convenience versus price sensitivity)? Should we Similarly, a ‘full column’ implies the presence of a use some other measure of consumer response to segment with a deep level of attachment to the marketing variables (e.g. likelihood of buying a firm’s product line. new product concept, response to price promotion, These three different but complementary participation in a loyalty programme)? The ‘best approaches to the segmentation audit illustrate practice’ in this area suggests three propositions: important principles to keep in mind. First, effect- ive segmentation requires a good deal of effort and 1 An effective basis for segmentation should allow attention. Cognitive effort, financial resources and one to differentiate among segments based on time commitment from top management are pre- their response to marketing variables; thus buyers requisites to the development of a viable segmen- versus non-buyers or price sensitive versus non- tation strategy. A firm with limited commitment is price sensitive are possible bases for segmentation. unlikely to simply happen upon an effective strat- Age, sex, marital status, psychographic egy. Second, all good segmentation approaches characteristics or other general characteristics of seek to ‘triangulate’ – or converge on understand- the consumer may not be good bases for ing segments through a different lens (description, segmentation since they do not assure differential attitude, behaviour, movement, value). Third, seg- response to marketing variables. mentation and the product/service portfolio must 2 The selected basis for segmentation should be reflect and reinforce each other. directly related to the strategic purpose of the segmentation effort. In general, there are two types of segmentation with two different Decisions required in implementing underlying rationales: A general segmentation of the market which a segmentation strategy allows the organization to organize itself around the selected target segments. As an While poor segmentation can result from flawed increasing number of companies shift from a thinking or conceptions of segmentation, it is our product management organization to a experience that many segmentation efforts fall flat market-driven organization or a matrix because of poor execution and implementation. organization of product by market (as might Ch11-H8566.qxd 8/8/07 2:04 PM Page 227

Market segmentation 227

be implied by the PSM analysis), it is critical to needs, brand loyalty and switching patterns identify relatively stable and large segments or a hybrid of the above variables. which could serve as strategic business units. 3 To gain a better understanding of the various Examples of such segments, in the case of a segments and their characteristics, it is critical to financial service firm, are the ‘Delegators’ – profile the segment’s key discriminating individuals who prefer to have a money characteristics.These include the complete manager take complete control over the segment profile on demographic, psychographic, management of their financial affairs, and the product usage, perceptions and preferences, ‘Electronic DIY’ – who prefer to do it attitudes and the like. In the case of industrial themselves using direct computer trading. To (business-to-business) segmentation, these variables reach these two segments effectively the firm should include both the characteristics of the needs distinctly different strategies. Members organization and each of the key members of the of each of these strategic segments share relevant buying centre.Table 11.2 identifies a list of some common financial/investment needs, variables commonly used as a basis for yet each of these segments may still be quite segmentation and as segment descriptors. It is heterogeneous with respect to other needs important to note that variables not used as a basis and, thus, could benefit from further sub- for segmentation can become descriptors of segmentation into more homogeneous segments that are developed based on other bases. groups. Specific segments for specific marketing and Selecting a research programme business decisions. For example, in the introduction of a new product, this may mean The quality of a segmentation programme depends a focus on the segment that has the highest largely on the quality of information used in likelihood of buying the product. In the launch developing the scheme. Segmenting any market of a new online electronic shopping service, it requires information on the characteristics of the could involve a focus on time-constrained market including: individuals with high-speed Internet access at home. Other specific segments and their Attractiveness:The size and growth of the market associated characteristics can be developed segment(s). for each of the marketing mix decisions. Decision making:The evaluation and choice criteria Notice that in the following examples the used by individuals (or buying centres in business- informational requirements (and therefore to-business marketing contexts), including the basis for segmentation) differs systematically benefits they seek and problems they try to solve. from issue to issue: Perceptions and preferences:The perceptions of For positioning: product usage, product preference for attitudes towards and usage of the preference, benefits sought or a hybrid of products and services of the firm and its the variables above. competitors. For new product concepts (and new product Characteristics:The demographic characteristics of introduction): intention to buy, preference the segment members or other relevant stable over current brand, benefits sought. characteristics of the buying centre in the case of For pricing decisions: price sensitivity, deal business-to-business markets. proneness, price sensitivity by Attitudes/feelings:The psychographic profile purchase/usage patterns. (lifestyle, personality and other psychological and For advertising decisions: benefits sought, attitudinal characteristics). media usage, psychographic/lifestyle. A Responsiveness: Reaction and sensitivity to the hybrid of the variables above with or marketing actions of the firm and its competitors. without purchase/usage patterns. For distribution decisions: store loyalty and To collect data on these variables and to analyse patronage, benefits sought in store and interpret them requires a systematic research selection. programme. Historically, segmentation research For general understanding of a market: centred on customer surveys. Yet there are a num- benefits sought, or in industrial markets, ber of additional approaches that should be con- the criterion used is purchase decision, sidered. Figure 11.2 outlines the range of options. product purchase and usage patterns, Formal primary research and especially surveys on Ch11-H8566.qxd 8/8/07 2:04 PM Page 228

228 The Marketing Book

Table 11.2 Variables commonly used as basis for segmentation and as descriptors of segments

Basis for segmentation Descriptors of segments

Organizational External Share Socioeconomic, political environment (culture, Trial technology, economic, political, regulatory, legal) Purchase/adoption Behaviour towards competitors Source loyalty Etc. Price sensitivity Organizational Customers of key competitors Industry type (e.g. SIC) Etc. Size Buying Centre Degree of centralization Buying process Capabilities (technical, financial, etc.) Informational search Geographic location (country, region, city, etc.) Criteria/benefits sought Etc. Negotiation style Buying centre Application Size Decision Composition Post-purchase evaluation Buying situation Etc. Influence Consensus among the members Individual Buying process Awareness Buying organization and policies Knowledge Relations with suppliers Perceptions, preferences, attitudes towards Etc. the brand Preference Individual Recommendation Demographic (age, sex, family life cycle, Purchase income, education, social class, etc.) Usage Psychographics (personality, lifestyle, activities, Loyalty interest opinions, etc.) Etc. Etc.

a sample of customers and prospects have been the could yield insights into the characteristics of vari- most common approach to segmentation research. ous market segments and their response to mar- Recent developments of databases on the entire keting activities. One of the most important universe have changed the nature of segmentation developments of recent years is the adaptation of research. Consider, for example, the pharmaceutical traditional marketing research methods for appli- industry. Here a number of firms are developing cation on the Internet. Dahan and Srinivasan databases on the entire industry. These databases (2000), for example, show which conjoint analysis include hundreds of thousands of physicians, replicates on the Internet. Of equal importance is thousands of hospitals and thousands of managed the work by Dahan and Hauser (2002) that docu- care organizations. ments new research methods which are facilitated A second type of formal research includes sec- by the unique communication capabilities afforded ondary data analysis using geo-demographic data by the Internet. which can be used both for segmentation research The third approach is input from ongoing as well as for experimentation. Adaptive experi- business activities. In this case, we include both mentation (Wind, 2007) is another approach which data on the market response to various strategies Ch11-H8566.qxd 8/8/07 2:04 PM Page 229

Market segmentation 229

Problem definition and segmentation model

Input from ongoing Adaptive experimentation Formal research business activities

From strategic From the Primary research- Secondary alliances with regular business original data data sources end users activities sources

Sample Universe

Database, analysis, DSS, planning simulation, resource allocation model and/or a management subjective model using approaches such as the analytical hierarchy process (AHP).

Figure 11.2 Selecting a segmentation research programme

and insights gained from strategic alliances with Relative Attractiveness:What is the size of the customers. This latter approach is especially critical segment in terms of the revenues and profits it is in the new product development area when the likely to generate? firm works closely with its customers on the devel- Responsiveness:What is the likely response to the opment of new products and service. “Best research firm’s offering, including response to the practices” would include using a number of these positioning of the firm’s products and services? approaches, integrating the resulting data in a data- Competencies: Do we have the required offerings base, continuously updating the database and hav- and competencies to effectively develop, reach ing it as part of a decision-support system (DSS). and serve the selected segments? Exogenous factors:What is the likely impact of changes in relevant environmental conditions (e.g. Selecting target segments economic conditions, lifestyle, regulations, etc.) on Having segmented a market it is often not desirable the potential response of the target market or possible to pursue all segments. Thus, a critical segment? decision is the selection of target market segments. Accessibility: Can the segment be reached (via If more than one segment is to be addressed, then controlled media and distribution or via a self- consideration needs to be given to the order in selection strategy)? which each will be targeted. Moreover, all segments Cost:What are the likely costs of effectively have to meet four conditions: measurability (ability reaching the target segment? to measure the size and characteristics of the seg- Competition:What are the current and likely ment); substantiality (having a minimum profitable competitive activities directed at the target size); accessibility (ability to reach and serve the segment? segments) and actionability (ability to implement Portfolio management: How many segments can be strategies to serve the segments) (Kotler and Keller, managed effectively? What synergies or 2005). Beyond these fundamental requirements, the relationships exist among the segments? selection of target market segments requires answers to the questions given below. (Note the These and related criteria often require information relationship between these questions for selecting that is not always available from market research. segments, and those for segmenting the market In addition, it is important that any information itself, given previously.) collected through marketing research or other Ch11-H8566.qxd 8/8/07 2:04 PM Page 230

230 The Marketing Book

Well-being of the firm

Short term Long term

.80 .20 Planning horizon:

Reduce Profit Profit growth Sales growth downside risk .39 .34 .20 Objectives: .07

Segment attractiveness Strength in segment Synergy

.20 .51 .21 Criteria:

Current segments New segments

Market A B C D E F G H* I* J* K* L* segments: .10 .21 .11 .02 .03 .18 .02 .10 .17 .01 .03 .02

Note: The numbers are the composite priorities of each item. Figure 11.3 An illustrative output of an AHP designed to select a portfolio of market segments

sources is structured in a way that is amenable to The segment selection procedure illustrated in action. Consequently, a decision-support frame- Figure 11.3 utilized the analytic hierarchy process work is needed to aid managerial decision making. (AHP). The AHP (Wind and Saaty, 1980; Saaty, This framework should utilize both managerial 1992) is a measurement approach and process that experience and available data. At the minimum it helps quantify management subjective judge- should include: identifying relevant criteria, and ments. The essential steps in implementing an evaluating the segments on the selected criteria. AHP include: (1) setting up the decision problem Following the logic of product and business by the relevant management group as a hierarchy portfolio analysis, a portfolio of current and poten- of interrelated decision elements, (2) evaluating the tial market segments can be constructed (Wind and various elements of the hierarchy by pair-wise Robertson, 1983). Figure 11.3 illustrates a portfolio comparisons and (3) using a mathematical method of segments in which each segment is evaluated on to estimate the relative weights of decision elem- its attractiveness, on the firm’s expected strength in ents. The Decision Lens (http://www.decisionlens. the segment and on synergy. The first two criteria com) process and software greatly facilitates the are the same dimensions used in the GE/McKinsey implementation of an AHP and its extension to the portfolio matrix. These dimensions can be based on Analytic Network Process (ANP). a single criterion or represent a composite of mult- The application of AHP or ANP to segment iple criteria. For example, segment attractiveness selection involves bringing together the key execu- could include such factors as the segment’s size and tives and presenting to them all available market- growth, risk factors and the cost of reaching the seg- ing research information. This provides input to ment. Segment strength could include such factors the deliberation and evaluation of the various as current and expected share in the segment and segments, in a pair-wise comparison, against expected profitability. The specific criteria and their each of the chosen criteria. The results for the pri- relative weights can be determined by management oritization of the segments from our illustrative judgement and marketing research input. example are given in Figure 11.3. An examination Ch11-H8566.qxd 8/8/07 2:04 PM Page 231

Market segmentation 231

of this illustrative hierarchy suggests a number of section, AHP was utilized to select a portfolio of seg- segmentation-related conclusions: ments. It also provided guidelines for resource allo- cation among the segments. Allocation of resources Management established three criteria (segment typically involves not only the allocation among attractiveness, strength in segment and synergy) segments, but also the allocation of resources to the which vary in their importance with respect to marketing mix variables – product, price, distribu- the firm’s ability to achieve their four objectives tion, promotion and advertising. (profit, profit growth, sales growth and downside The basic problem of resource allocation is to risk).The objectives, in turn, vary in their decide on the mix of resources that generates opti- importance under short- or long-term conditions mal response (sales, profitability, etc.). Estimating (not shown in the figure).The overall importance and modelling the sales response of each segment to of the four objectives, assuming an 80–20 the various marketing resources is critical. How, for weight for short versus long term is presented in example, should sales persons currently covering a Figure 11.3.The seven current segments when market be allocated across segments to optimize evaluated against the three criteria (segment their return? A comprehensive set of analytical attractiveness, strength and synergy), which in tools for how to answer such questions have been turn are weighted by their importance to the developed in Lilien and Rangaswamy (2002). Other accomplishment of the four objectives (weighted traditional approaches, typically based on conjoint by their importance for the short- and long-term analysis studies among current and potential cus- well-being of the firm), suggest that three of the tomers, have been applied to a wide range of situ- segments – D, E and G – are not very attractive ations including computers, telecommunications and should be considered as candidates for products and services, pharmaceuticals, etc. (Green deletion, or at least destined to receive no and Krieger, 1985; Krieger et al., 2004). In the cases incremental resources. in which empirically based market response func- Five new segments were identified.When tions are not available, management’s subjective evaluated on the three criteria, two of the judgement, using either decision calculus methods segments – H and I – were viewed as candidates (Little, 1979) or the AHP (Wind and Saaty, 1980), for inclusion in the portfolio and three – j, K and can be employed. Above all else, these approaches L – as candidates for exclusion. stress the importance of structure and managerial As a result of the process, a new portfolio of input. The principles underlying successful adop- segments was established with segments A, B, C, F, tion and execution of marketing science models are H and I. clearly articulated by Lodish (2001) aptly titled The outcome also suggests how resources should Building Marketing Models that Make Money. The be allocated to each segment. Since the complexity and importance of the resource dimensions included in the hierarchy encompass allocation decision suggests the advisability of both the expected benefits from each segment as employing a methodology such as the AHP that well as the cost of reaching them and risk, the incorporates managerial judgement, empirical data priorities can be used as a rough guide for derived from econometric market response models resource allocation.This would lead to the and experiments, consumer studies and decisions following allocation:A 11 per cent, B 24 per models. cent, C 13 per cent, F 21 per cent, H 11 per cent and I 20 per cent. Implementing a segmentation strategy The most difficult aspect of any segmentation This example reinforces the idea that the identifi- project is the translation of the study results into cation and selection of market segments not only marketing strategy and programmes. No rules requires explicit structure, but also quantification can be offered to assure a successful translation and prioritization wherever possible. and, in fact, little is known (in the published litera- ture) on how this translation occurs. However, Allocating resources among segments informal discussions with executives and obser- vations from our own experience of ‘successful’ The selection of target segments and the allocation and ‘unsuccessful’ translations suggest a few of resources among segments are interrelated and general conclusions on the conditions likely to iterative decisions. For example, in the preceding increase successful implementation. Ch11-H8566.qxd 8/8/07 2:04 PM Page 232

232 The Marketing Book

1 Involvement:All relevant decision makers (e.g. includes those cases in which a priori segmentation product managers, new product developers, design such as product purchase or usage, loyalty advertising agency, sales management, etc.) must or customer type is augmented by some cluster- be involved in the problem definition, research based segmentation on variables such as benefits design, strategy generation and evaluation. sought. Hybrid designs are especially common in 2 Texture:The segment characteristics used in the business-to-business segmentation. analysis should be rich enough to provide the Approaches such as macro and micro segmen- basis for innovative product positioning and tation (Wind and Cardozo, 1974) and sequential product, pricing, communication and distribution clustering (Peterson, 1974) are often used in prac- decisions. tice. Sequential clustering, for example, clusters on 3 Direction:The segment characteristics should also some market-based demographics followed by atti- provide guidelines for the generation of creative tude (or benefit) clustering within each demo- executions of the selected strategy. graphic segment. In both cluster-based design and 4 Ownership: Given the oftentimes non-marketing hybrid design, the size and other characteristics orientation of product management, it can be (demographics, socioeconomic, purchase, etc.) of desirable to shift the responsibilities for the segments are estimated. Classification proced- segmentation strategy to segment-driven ures based on cluster analysis and variants thereof managers and teams. are especially useful in providing management 5 Science: Since vital information on the cause and with a data-driven view of segmentation. A ‘classic’ effect relationship between marketing resource illustration of this value is given in Moriarty and allocation and response is often absent, it is vital Reibstein (1986) who use clustering methods to that firms undertake to explicitly measure and reveal a view of the market segments of the per- monitor outcomes. For example, linking market sonal computer market that challenged manage- response to media and distribution systems. ment intuition about the way the market was thought to be structured. In particular, they are able to show how benefit segmentation relates to firm descriptors such as size and industry classification. Advances in segmentation research Discrimination methods Many of the papers published by academics and practitioners in Marketing Science, Management Discrimination methods employed to establish the Science and the Journal of Marketing Research offer profile of the segments commonly use multiple specific advances in segmentation research and discriminant analyses or regression analysis. Such modelling. Moreover, Interfaces offers in-depth case statistical methods are often augmented with studies of marketing science tools that demon- graphic packages that graphically display the pro- strably improve the bottom line for the client firm. file of the segments. These methods are especially While the approaches outlined in these articles are useful when the goal is to assign new individuals necessarily diverse and sometimes focused in their (e.g. ‘Prospects’) into groups based on their likely application, they nevertheless are centred around disposition to use the firm’s products and services. six sets of key segmentation ‘tools’. We list the tools In recent years, discriminant-based techniques below in approximate increasing order have fallen out of favour due to an absence of effective solutions to implementation and inter- pretation problems such as those documented in Classification methods Eisenbeis (1977). Classification methods for segmenting markets are especially critical for clustering-based segmentation approaches and for hybrid designs. A clustering- Simultaneous evaluation based segmentation design involves determining the One area receiving more attention from academic number, size and characteristics of the segments researchers involves application of sophisticated based on the results of clustering of respondents on multidimensional scaling and multivariate statis- a set of ‘relevant’ variables. Benefit, need and atti- tical analyses in order to jointly locate segments and tude segmentation are examples of this type of products in a consistent market map. One early approach. Significant advances have been made in development was the componential segmentation the clustering area. Hybrid segmentation design approach proposed by Green and DeSarbo (1979). Ch11-H8566.qxd 8/8/07 2:04 PM Page 233

Market segmentation 233

This method shifts the emphasis of the segmenta- respondent segment the frequency of first choices tion model from the partitioning of a market to a for each of the new product combinations con- prediction of which person type (described by a sidered, and (2) for each new product combination, particular set of demographic and other psycho- the frequency of first choices across segments. graphic attribute levels) will be most responsive to Componential segmentation offers a new conceptu- what type of product feature. Componential seg- alization of market segmentation in that it focuses mentation is a logical extension of conjoint analy- on the building blocks of segments and offers sis and orthogonal arrays to cover not only simultaneously an analysis of the market segment product features but also person characteristics. In for any given product offering and an evaluation of componential segmentation the researcher is inter- the most desirable product offering (or positioning) ested in parameter values for various respondent for any given segment. The concept and algorithm characteristics (demographic, product usage, etc.) of componential segmentation can be extended to in addition to those for the product stimuli. cover not only two data sets (product feature and In a typical conjoint analysis approach to mar- respondent characteristics) but three or more data ket segmentation, a matrix of subjects by utilities is sets by adding, for example, the components of developed. This matrix can serve as the input to usage situations and distribution options. the determination of the profile of some a priori Other studies have used the output of logit segments (e.g. product users versus non-users) or and nested logit choice models to the simultan- alternatively as the input to a clustering pro- eous estimation of segment size, characteristics and gramme which would result in a number of bene- choice probabilities to uncover market structure. fit segments. In componential segmentation, the Examples of this work can be seen in Bucklin and same design principles which guide the selection Gupta (1992) and other approaches documented of (product) stimuli are applied also to the selec- in Grover and Srinivasan (1987, 1992). Elrod tion of respondents. For example, in a study for a and Keane (1995) use panel data to simultan- new health insurance product, four sets of respond- eously uncover latent attributes and consumer ent characteristics were identified on the basis of preferences. More recently, Moe and Fader (2001) previous experience and management judgement: use a stochastic modelling approach which allows age, sex, marital status and current insurance sta- clusters of products to draw from a fixed popula- tus. Employing an orthogonal array design, one tion of underlying segments. One advantage of screens respondents to select those who meet the this approach is that it simultaneously considers specific profiles specified by the design. Each rates of purchase within segments and the mix of respondent is then interviewed and administered segments that interact with different clusters of the conjoint analysis task for the evaluation of a set products. of hypothetical health insurance products (also A conceptually related, but methodologically selected following an orthogonal array design). different stream of research has sought under the Having completed the data collection phase, decision rules used by members of different mar- the researcher would have a matrix of averaged ket segments. Kamakura et al. (1996) illustrate that profile evaluations of the product stimuli by the segments are not only heterogeneous in prefer- groups of respondents. This data matrix is then ences, but also in information processing and con- submitted to any number of componential seg- sideration of product attributes. In that study, mentation models, which decompose the matrix members of the ‘Brand-type’ segment focus first into separate parameter values (utilities) for each on brand, then on product form; ‘Form-type’ seg- of the levels of the product feature factors (com- ment members reverse this process. Gilbride and prising the stimulus cards) and separate param- Allenby (2004) show how to uncover and test for eter values (saliences) for each of the levels of the non-compensatory and compensatory decision customer profile characteristics (describing the rules using only choice data. respondents) which indicate how much each pro- file characteristic contributes to variation in the evaluative responses. Simulation and optimization Given these two sets of parameters, the Simulation and optimization are at the core of researcher can make predictions about the relative models for selection of a target portfolio of seg- evaluation of any of the possible product features ments. These models are typically based on conjoint by any of the respondent types. The results are analysis studies. Among the most powerful of these used with a simulation to estimate: (1) for each approaches is flexible segmentation. In contrast to Ch11-H8566.qxd 8/8/07 2:04 PM Page 234

234 The Marketing Book

a priori segmentation in which the segments are determined by a series of multivariate statistical determined at the outset of the study and cluster- analyses which can be incorporated into the simula- ing-based segmentation in which the selected seg- tion. Some of the more significant developments in ments are based on the results of the clustering the segmentation area in the last two decades began analysis, the flexible segmentation model offers a with advances in simulation and optimization pro- dynamic approach to the problem. Using this cedures and associated user-friendly software. approach, one can develop and examine a large Early developments by Green and Krieger (1991, number of alternative segments, each composed of 1994) greatly facilitate the task of selecting an those consumers or organizations exhibiting a simi- optimal (or close to optimal) set of segments, lar pattern of responses to new ‘test’ products and have now been incorporated into Adaptive (defined as a specific product feature configur- Conjoint Analysis (ACA) routinely implemented by ation). The flexible segmentation approach is based Sawtooth Software, one of the leading industry sup- on the integration of the results of a conjoint analy- pliers in the US (see e.g. Bryan Orme’s Introduction sis study and a computer simulation of consumer to Market Simulators for Conjoint Analysis). A rigor- choice behaviour. ous practitioner-oriented account of recent develop- The simulation model in a flexible segmenta- ments and applications can also be found in Krieger tion approach uses three data sets: et al. (2004) and Green et al. (2003). One of the more promising lines of research 1 Utilities for the various factors and levels for each in simulation and optimization is emerging from respondent. the joint efforts of researchers in operations 2 Perceptual ranking or rating of the current research and marketing. Camm et al. (2006) provide brands on the same set of attributes. an exact algorithm to identify the new product 3 A set of demographics and other background concept that will maximize the number of respon- characteristics. dents for whom that product exceeds a particular utility threshold. This is an important develop- The active participation of management is also ment as prior research in marketing was only able required to design a set of ‘new product offerings’ to identify heuristics for this task and therefore (each defined as a unique combination of product could not guarantee global optimality. features – specific levels on each factor included in the conjoint analysis study). Management par- ticipation can be on a real-time basis in which Bayesian methods managers interact directly with the computer simu- One important development in segmentation lation. Research suggests that cross-functional research is the application of Bayesian methods to teams should also have input into the conjoint simu- segmentation problems. Bayesian methods allow lator approach. For example, Srinivasan et al. researchers to estimate individual-level response (1997) leverage inputs from marketing, design and parameters and therefore connect the modelling manufacturing in order to offer ‘customer-ready’ approach to the conceptual notion of ‘one-to-one’ prototypes to respondents. or ‘customized’ marketing. These methods have The choice simulator is based on the assump- become increasingly prevalent in a host of segmen- tion that consumers choose the offering with the tation-related studies that cover issues ranging highest utility and is designed to establish the con- from coupon response (Rossi et al., 1996), advertis- sumer’s share of choices among the existing brands, ing effectiveness (Schweidel et al., 2006), price which can be validated against current market response and new product design (Sandor and share data if available; and the consumer’s likely Wedel, 2005; Evgenio et al., 2005). Review articles switching behaviour upon the introduction of any on the impact of Bayesian methods and the new product. This phase provides a series of brand- implications for market segmentation are given in switching matrices. Within each matrix manage- Allenby and Rossi (1999) and Rossi and Allenby ment can select any cell or combination of cells as a (2003). possible market segment (e.g. those consumers remaining with brand i versus those who switched to new brand, etc.). Once the desired segments (cell Customer-to-customer interaction or cell combination) have been selected, the demo- One fundamental shift in the practice of segmen- graphic, lifestyle, product purchase and usage, tation involves the harnessing of ‘customer-to- and other relevant segment characteristics can be customer’ interactions for segment involvement Ch11-H8566.qxd 8/8/07 2:04 PM Page 235

Market segmentation 235

and development. Customer-to-customer intera- Too narrow a focus ctions facilitate the evolution of new segments and This criticism encompasses five areas that can be the transmission of product-relevant information addressed using specific advances in modelling and preference data. These interactions are explicit technologies: in evolving social networks (e.g., MySpace, YouTube, LinkedIn) and will be increasingly important to firms of all types. Consider, for 1 The traditional focus on ‘one segment per customer’ example, 1800diapers.com which has grown from (the assignment of individuals to mutually a start-up with no customers to over 100,000 regis- exclusive and collectively exhaustive segments) is tered customers in the space of a few months. too narrow.This problem can be overcome by the A particularly important segment for this business use of overlapping clustering using a clustering is the cluster of customers who respond to the procedure that allows for an individual to belong incentive to recruit other customers. This notion of to more than one segment. For a discussion of employing existing customers and segments to this method, see Arabie et al. (1981) and create new ones is analysed formally using data- Chaturvedi et al. (1994). One could also argue that mining methods (see subsequent section in this it is more appropriate to jointly model product- chapter) by Hill et al. (2006) in their application to segment relationships as Moe and Fader (2001). telecommunications data. Among other things, the Moreover, the interrelationships among product authors find that customers linked to a prior cus- purchases across multiple product categories also tomer adopt new services of the firm at a rate that provide insight into which customers are most is three to five times higher than baseline groups worth pursuing. For an application of this method that were selected according to the firm’s existing see Kamakura et al. (2004). best practices. The relationship between physical 2 One segmentation scheme fits all.This problem of and virtual proximity is also likely to influence trying to fit one segmentation scheme for all customer acquisition and segmentation strategies marketing decisions can be solved by employing for firms that have spatially dispersed clienteles. In the flexible segmentation approaches or by examining trials of a new online grocery service, developing a number of segmentation schemes Bell and Song (2007) find that ‘neighbourhood and linking them. Moreover, the results from effects’ play an important role in generating new multiple schemes should be complementary and customers, even after controlling for typical demo- convergent.While traditional segmentation graphic and other location specific variables. approaches focus on uncovering differences in behaviours and preferences, it may also be Linking segmentation findings with important to consider other possible segmentation approaches.This includes variation management subjective judgements in decision-making rules or heuristics adopted by Given the complexity of deciding on the portfolio of customers (see Gilbride and Allenby, 2006). segments to target, it is helpful to use a framework 3 Neglect of sub-segments.With the exception of and methodology that captures managements sub- segments of one, most segments are jective judgement while allowing the incorporation heterogeneous. It is important to recognize this of findings from various segmentation studies. The and augment the basic segmentation with AHP, as illustrated in Figure 11.3, and the more gen- additional sub-segmentation.This is the concept eral ANP is ideally suited for this task. underlying hybrid segmentation models and is increasing in its popularity. In this context, it is Addressing the problems also helpful to develop a hierarchy of segments. A byproduct of this research is the development In addition to methodological advances in these of measures on the segmentability of each market seven areas, some of the more interesting advances (and the degree of homogeneity of selected in segmentation research are those developed to segments). Several methods have been developed address three of the criticisms of segmentation to deal with within-segment heterogeneity.These research. Namely, that it: (1) has too narrow a focus, include the multi-mode Bayesian methods of (2) is static and deterministic and (3) is poorly inte- Allenby et al. (1998) and the empirical Bayes grated with strategy. The advances in addressing approach of Kamakura and Wedel (2004). these problems are briefly discussed in the follow- 4 Individual as the unit of analysis. Most segmentation ing section. studies use data on individuals. However, few Ch11-H8566.qxd 8/8/07 2:04 PM Page 236

236 The Marketing Book

decisions are made by a single individual. Most Static and deterministic perspective households and industrial (business-to-business) A major limitation of many of the segmentation decisions are made by a buying centre.An studies is their neglect of the dynamic aspects important advance in segmentation research is the of segmentation. Static/deterministic segmenta- shift in the unit of analysis from individuals to tion tends to ignore segment change and market buying centres, which may be a pair of individuals dynamics; ignoring such changes often has several in a family unit, or a more complex set of consequences. Johnson and Lilien (1994) provide a relationships within an organization. Recognizing conceptualization and model-based approach to the heterogeneity within a buying centre, the use deal with segment dynamics. The Internet is one of key informants as representative of the buying environment in particular that requires special centre is often not appropriate.A better approach attention to dynamics and several authors have would be to identify all the members of the addressed this issue. A recent example is given in buying centre and collect the information on a Reutterer et al. (2006) in which the authors develop subset of the buying centre members.This allows a dynamic model built from purchase history an assessment of the level of consensus among information provided by a loyalty programme. In members of the buying centres.The level of general, to address the issue of dynamics one can consensus as well as the composition of the do the following: buying centres can be used as a basis for, or descriptors of, segments. Choffray and Lilien Define bases for segmentation that focus on (1978) demonstrate critical differences in decision change. criteria among members of buying centres for Monitor changes in segment composition over time. industrial heating and cooling equipment. In Focus on strategies that will change segment addition,Wilson et al. (1991) describe and test the membership (from non-users to users, light to best ways to combine the preferences of heavy users). individuals and buying centre members when Incorporate competitive actions and reactions, since trying to determine how the buying centre is the desirability of a segment depends not only on most likely to act. Individual preference formation the segment’s characteristics and our own strategies and the integration of preferences in a group but also on competitive actions and reactions. process are analysed in Arora and Allenby (1999). The authors show that the method also facilitates A second major weakness of much of the segmenta- revelation of ‘high influence individuals’ as part of tion research is the missing stochastic component. the estimation and segmentation process.A more While early and seminal methods of segmentation recent paper by Aribarg et al. (2002) examines the (e.g. Kamakura and Russell, 1989) explicitly recog- separate processes of preference formation and nize that classification into segments is a probabilis- consensus building and shows how the two tic outcome, many applications seem to gloss over interplay to produce final decisions. Finally, state of this important fact. One new direction in this area the art work by Arora (2006) shows that one may addresses the ‘stochastic’ component of the seg- in fact impute joint preference using smaller and mentation problem directly by explicit introduction more cost effective datasets using ‘sub-sampling’. of estimates of segment size and characteristics. The A formal treatment of sub-sampling – or the academic research world has also examined fuzzy repeated sampling from observations which exhibit clustering (see, e.g. Wedel and Steenkamp, 1991; dependence – is given in Politis et al. (1999). Rayward-Smith, 2002), although this method has 5 The segmentation of the month. Segmentation area yet to be widely embraced and applied in the mar- has not escaped other management fads and has keting community. had its share of ‘segment of the month’ promise and advocates.To avoid this trap, one must attempt theory-driven segmentation. It also helps Poor integration with strategy to keep the focus on market response variation as To address the problem that segmentation studies bases for segmentation and to include all other are often not reflected in the resulting strategy, a variables as segment descriptors.A recent number of actions can be taken. These include: conceptual article on ‘marketing malpractice’ illustrates that segmentation schemes can become 1 Analysis. Carefully map the results of all studies cumbersome, a-theoretic and consequently, such as copy, concept, product, distribution and ineffectual (see Christensen et al., 2005). other marketing studies at the segment level.To Ch11-H8566.qxd 8/8/07 2:04 PM Page 237

Market segmentation 237

the extend possible, develop empirical and their linkage to decision-(and executive-) generalizations regarding your target segments. support systems (DSS/ESS) and dashboards, which 2 Targeting.Avoid infrequent and expensive large in turn can include expert/knowledge-based sys- base-line segmentation studies and instead tems. In this context, many advances in segmenta- include in all marketing and business studies a tion research and modelling can be incorporated in segmentation analysis. the DSS; moreover, much of our knowledge of seg- 3 Linking. Link the segmentation to positioning and mentation can be developed as rules for a know- its associated marketing mix strategies. Specifically ledge-based system that could help management to recognize the interdependence between the two. select target segments. Given a positioning, what is the best segment(s)? Some of the advances in this area include the Given a segment, what is the best positioning(s)? ability to have ‘live’ databases in which one can Having fixed on a segment/positioning, it is critical update on an ongoing basis the customer data- to develop a marketing strategy that will meet the base. Consider, for example, Citibank’s interactive needs of the selected segment and reflect the intelligent DSS which guides all interactions with target positioning. the customer. These interactions include the deliv- 4 Implementation. Carry the segmentation efforts to ery of a direct mail or telephone sales message tar- the sales force level by encouraging each sales geted by the system which also coordinates a person to segment his/her market. In addition, number of customer interactions. The coordinated segmentation methods can also be very helpful in interactions create a ‘dialogue’ to follow the con- segmenting the sales force (or other aspects of sumer with the appropriate intervention at each the distribution channel). touch-point of contact of the consumer with 5 Selection. Carry through the segmentation Citibank. The touch points include subsequent strategies to the business and corporate level by telephone enquiries, ATM use, bank teller inter- focusing on a portfolio of segments and by using actions and receipt of statements. This system is the portfolio of segments as the core of the based on a new relationship model with house- business and corporate strategy. hold and not only on the traditional banking focus on accounts.

Segmentation in the global Data mining information age Data mining (e.g. Lewinson, 1994) offers enormous potential for empirically driven insights. ‘Data The information revolution has been the subject of mining’ refers to a number of pattern recognition an increasing number of scholarly studies that has models that may use neural nets or fractal technol- captivated the imagination of scholars, managers ogy to discover patterns in the data. These methods and the population at large. This revolution is have been employed commercially to identify seg- greatly affecting the ways in which firms are man- ments of customers most likely to buy a given aging their operations and is likely to change dra- product (in banking, e.g.) and to the identity of cus- matically the way in which business is conducted. tomers most likely to leave (cellular telephones, Information strategy is at the heart of most recent e.g.). These approaches, although intriguing, are efforts to reengineer and reinvent the corporation still in their infancy and require further validation. and is leading to the creation of a new management Some marketing researchers have combined data- and associated marketing paradigm. mining methodologies with more traditional statis- In the new management and marketing para- tical models. Cooper and Giuffrida (2000), for digm (Wind, 2005) information is having a pro- example, utilize data mine the residuals of a trad- found effect on: (a) the nature and quality of itional segment-based model of response to promo- management decisions; (b) the nature of business tions and use this combined model to improve strategies and (c) the creation of innovative com- sales forecasts. A notable exception in this area is munication and distribution systems. the work of Levin and Zahavi (2001) who compare several segmentation methods used in a predictive modelling context – several supervised-learning Management decisions segmentation methods involving decision trees, Management decisions are affected by the availabil- and a couple non-supervised methods involving ity and use of databases on the entire population judgemental FRM and FRAT methods. This line of Ch11-H8566.qxd 8/8/07 2:04 PM Page 238

238 The Marketing Book

work is promising and adds power to segmenta- Rogers (1993) presented a vision of their one- tion research and modelling. to-one paradigm which includes and focuses on: Whereas the Citibank example illustrates future development, much of the work today share of customers not share of market; relies more on the traditional geo-demographic collaborating with customers to create products segmentation based on consensus and other data. and relationships; An example of this type of effort is the Claritas customizing products, services and promotional Prizm lifestyle segmentation. This segmentation efforts for each customer; divides the USA into 62 clusters. These clusters economics of scope; are further divided into 15 groups that vary with engaging the customers in dialogues – the respect to the type of location – rural, town, interactive individualization of media is here. suburb or urban – and with respect to level of affluence. The Prizm lifestyle clusters can be The shift to segments of one requires a rethinking related to any target group of interest. Geo- of the segmentation concept as well as the develop- demographic information continues to generate ment and utilization of sophisticated databases, new applications and has been shown to be espe- marketing analysis, modelling and strategy. The cially useful in customer targeting (see e.g. Sleight, trend towards such developments is inevitable and 2004). Much of this work also signals a new inter- is accompanied by another discontinuous trend – face between geography and marketing and how the globalization of business. The early vision data sources and concepts can be combined for exemplified in the work of Peppers and Rogers has more effective segmentation and targeting (Harris recently been augmented by the notion of the et al., 2005). ‘Long Tail’ – a phrase coined by Chris Anderson and popularized in his 2006 book. In the long tail a multitude of ‘low demand’ products are targeted at Information strategies small segments of customers – such that their col- Information strategies are relatively recent add- lective share outweighs that of traditional ‘block- itions to the traditional set of marketing strategies. buster’ products. Effective segmentation is at the One of the early and most effective information- heart of delivering in long tail markets. based strategies was the ‘capture the client’ strat- egy of America Hospital Supply’s direct link Globalization between hospital computers and AHS computers, eliminating the need for human interaction in the Increasing numbers of industries are global. To suc- straight rebuy case. ‘Capture the clients’ strategies, ceed in this environment, firms have to shift from a such as the direct computer link between P&G and domestic perspective to considering the world as Walmart, are increasing in popularity. the arena of operations both with respect to the con- sumer markets for products and services as well as for the resources markets for raw material, R&D, New communication and distribution manufacturing, human and capital resources. The globalization of industries is also accom- options panied by trends towards regional economic inte- Information technology is also dramatically chang- gration – the European Union, NAFTA and the ing the nature of the communication and distribu- various other efforts for regional integration in Asia tion options. Electronic shopping developed at a and Latin America. The implication of these devel- much faster pace than was ever expected (see, e.g. opments for segmentation is that management has Blattberg et al., 1994; Rangaswamy and Wind, 1994) to consider portfolios of segments that include: and online business-to-consumer (B2C) sales in 2006 in the US alone were estimated at $212 billion global segments; (US Department of Commerce). regional segments; These changes affect all aspects of our lives segments within specific countries. and are altering the concept of segmentation. In the new marketing paradigm the traditional mass Added to this complexity is the need to consider as market is being replaced with segments includ- the unit of analysis not just countries but countries ing, at the extreme, segments of one. In a break- by mode of entry, since the risk and attractiveness through book, The One to One Future, Peppers and of a country depend on the mode of entry. The Ch11-H8566.qxd 8/8/07 2:04 PM Page 239

Market segmentation 239

selection and implementation of a portfolio of seg- cross-national segmentation schemes. A conceptual ments which includes global segments, regional overview of how to think about cross-national seg- segments and segments within countries (by mode mentation is given in Steenkamp and Ter Hofstede of entry) requires a significant amount of informa- (2002). Interestingly, very recent research by Ter tion on all relevant markets around the world. The Hofstede and Park (2007) suggests that spatial or creation and maintenance of such a data/know- geographic differences across countries provide ledge base is not a trivial undertaking and is one of considerable explanatory power for segmenta- the major obstacles to the development of global tion schemes. This result holds in an analysis of segmentation strategies. European Union consumers that controls for differ- Creation of processes for the development and ences in other critical variables including culture maintenance of country, regional and world data- and socioeconomic status. As global firms continue bases is a high priority undertaking for all global to struggle with the appropriate mix of ‘customiza- firms. Yet the development of effective segmenta- tion’ and ‘standardization’ there is enormous poten- tion can take place even without such databases tial for important conceptual and methodological if the firm will proceed in an iterative bottom–up breakthroughs in this area. and top–down segmentation. This process involves three bottom–up steps:

1 Segmentation of the market in each country (by Extending the segmentation concept mode of entry). 2 Examination of the resulting segments in all the In the marketing literature, in practice and in the selected countries to identify common segments discussion so far, segmentation has been limited across countries – clustering of country to ‘customer’ markets. Yet the concept applies to segments. all heterogeneous populations and can and 3 The creation of a global portfolio based on should be extended to the other stakeholders in various clusters of segments. the firm – all those who have a ‘stake’ in its sur- vival and growth. The resulting portfolio of segments should be com- Consider, for example, the firm’s own sales pared to a desired (top–down) conceptual portfolio force. Most large firms employ thousands of sales of segments. The comparison and contrast of the people who vary considerably in their perform- two portfolios should be driven by the concept of ance. The 20/80 rule often applies to them as it global operation which balances the need to does to the customers (i.e. 20 per cent of the sales develop strategies that best meet the needs of the force often accounts for 80 per cent of the profits). local markets (given the idiosyncratic market, com- In multi-product firms, different sales people often petitive and environmental conditions), while at tend to sell different mixes of products. They differ the same time trying to achieve economies of scale in their family life cycle stages and hence have dif- and scope by focusing on cross-country segments ferent financial and time needs (some are still wor- in a number of markets. The AHP framework and ried about college education for their kids while methodology can and has been used in this context others are single, etc.). These and other differences to help make such decisions, even in the absence of among the sales persons of any firm suggest that the needed ‘hard’ market data. As data become the traditional approach, in which a single sales available, both from the firm’s own experience and strategy and compensation is employed, is sub- from other sources, the data can be integrated in a optimal. To benefit fully from one’s sales force, it is database and used as input to the AHP process. critical to segment it. The segmentation of global markets offers The segmentation of the sales force based on enormous opportunities but is still one of the needs, benefits sought, expertise, perceptions and neglected areas of segmentation. It does offer preference or any other relevant characteristics intellectual and methodological challenges and is could lead to the identification of homogeneous critical from a management point of view. The lit- segments and the design of separate strategies erature presents scattered examples of global seg- towards them. In fact in any situation where man- mentation. Helsen et al. (1993) offer a proposed agement relies heavily on a sales force a dual mar- segmentation of multinational diffusion patterns, keting strategy should be developed – one for the while Ter Hofstede et al. (2002) extend this notion (target segments of) customers and a corresponding to include explicit modelling of spatial aspects of one for the (segments of the) sales force. Obviously, Ch11-H8566.qxd 8/8/07 2:04 PM Page 240

240 The Marketing Book

these two strategies should be coordinated and Of particular importance seems to be research integrated. Furthermore, a segmented strategy on the following areas: towards compensation is also desirable. To implement it while avoiding discriminatory prac- 1 Conceptualization. New conceptualizations of the tices requires the use of a compensation system segmentation problem. with a number of options relying on a self-selec- 2 Theory. Re-evaluation and operationalization of tion strategy in which the various sales people the normative theory of segmentation with could select the option most appropriate for their special emphasis on the question of how to needs. While the segmentation of the sales force allocate resources among markets and products and the resulting segmented strategies are likely over time. to meet with considerable resistance, future 3 New variables.The discovery and implementation research needs to address whether the benefits of new variables for use as bases for outweigh the difficulties and cost involved. segmentation (i.e. new attitudinal and behavioural Similarly, a segmentation strategy can bene- constructs such as consumption-based fit the firm’s dealings with its other stakeholders. personality inventories and variables which focus Wind (1978) described a segmentation of security on likely change in attitude and behavioural analysts and portfolio managers that led to better responses to the marketing variables) of the understanding by a firm of the criteria used in the markets for products, services and concepts. evaluation of firms in their industry and the per- 4 Research design.The development of new research ceptions of the given firm and its competitors. designs and parallel data collection and analysis Following a segmentation/positioning study, a techniques which place fewer demands on the strategy was developed to meet the needs of a respondents (i.e. data collection which is simpler target segment of security analysts that resulted for the respondent and data analysis procedures in a spectacular increase in the P/E ratio of the capable of handling missing data and incomplete given firm. block designs). Other stakeholders such as suppliers, cus- 5 Analytic methods.The development of simple and tomer service personnel, competitors, government flexible analytical approaches to data analysis agencies and the firm’s own shareholders are often capable of handling discrete and continuous heterogeneous. In all of these cases, understanding variables and selected interaction at a point in the key segments and selecting desired target seg- time and over time. ment(s) can greatly enhance the firm’s effective- 6 Boundary conditions. Evaluation of the conditions ness. In fact, as the cost of doing business in today’s under which various data analytical techniques environment increases, a segmented strategy may are most appropriate. be essential for any organization concerned with 7 Generalizations.The accumulation of knowledge the return of their marketing investments. on successful bases for segmentation across studies (product, situations and markets).This could entail meta-analytic work to identify success drivers for different bases and Issues and associated research agenda methodologies. 8 Validation. Undertaking external validation studies In one of the author’s introduction to the Journal to determine the performance of segmentation of Marketing Research special issue – segmentation strategies which were based on findings of research (Wind, 1978), the following conclusions segmentation studies. were presented: 9 Data generation techniques. Designing and implementing multi-trait, multi-method ‘Market segmentation has served as the focal point approaches to segmentation research aimed at for many of the major marketing research devel- both the generation of more generalizable opments and the marketing activities of most (reliable) and valid data. firms. Yet, too many segmentation researchers have 10 Cross-functional applications. Integration of settled on a fixed way of conducting segmenta- segmentation research with the marketing tion studies. This tendency for standardization of procedures is premature and undesirable. Given information system of the firm. the current state of the art, we offer the following 11 Translation. Exploring alternative approaches to 12-question areas as particularly ripe for new the translation of segmentation findings into undertakings’. market strategies. Ch11-H8566.qxd 8/8/07 2:04 PM Page 241

Market segmentation 241

12 Implementation. Studies of the organizational 3 Management and implementation. Development of design of firms which were successful and the processes and organizational architecture unsuccessful in implementing segmentation required to assure effective adaptation and strategies. management of segmentation.This includes adoption of the segmentation concept as the As noted in this chapter, numerous innovative foundation of all marketing and business decisions approaches to segmentation have evolved over the (as outlined in Figure 11.1) as well as the past 20 years. The centrality of the segmentation development of guidelines for effective concept within the marketing field is still para- management of segmentation. mount, yet further work on the new conceptual 4 Implicit segmentation.The effect of the Internet on and methodological aspects of segmentation should within and across customer information be undertaken. Review of some of the issues and transmission and demand aggregation is current advances in segmentation research indi- enormous (as acknowledged earlier in this cates that despite the great advances in the man- chapter). It is therefore critical that firms agement of and research practice of segmentation understand and utilize customer-to-customer numerous frontiers still require creative and sys- interactions as part of the segmentation process tematic study. Furthermore, despite the advances and allow customers to self-select and engage in in academic research during the past 30 years, we mutually beneficial patterns of influence (consider still do not have satisfactory solutions for many of the earlier example of 1800diapers.com). Such the issues raised in the 1978 research agenda approaches may be more effective and leveraged (Wind, 1978); yet new issues have emerged as than attempts to explicitly segment the market, well. The changes in the business environment and regardless of the quality of the underlying data especially the implications of operating in a global and statistical methods. information age, the emergence of empowered consumers and social networks do suggest, how- Addressing these newer challenges and continu- ever, the need to challenge our mental models ing to build our knowledge concerning the items of segmentation (Wind et al., 2004) and add a few identified in the 1978 research agenda is critical in additional items to the research agenda. These order to increase the value of segmentation. include: Continuous innovation and improvement in seg- mentation research and modelling for generating 1 Problem orientation. Reconceptualization of and evaluating segmentation strategies is neces- segmentation problems in light of the impact of sary, but not sufficient. Real progress in this area operating in the global information age and the requires rethinking the role of segmentation in the emergence of segments of one seems a useful global information age and concentrated efforts by first step. Many of the issues raised in this review management to develop and implement innova- from trends in customer-to-customer interaction tive and effective segmentation strategies. to the application of sophisticated Bayesian and The obstacles to effective segmentation are not data-mining methodologies suggest that strong methodological, nor even the lack of data, but conceptualization is still a necessary condition for rather the ability and willingness of management to effective segmentation. undertake a segmentation strategy and establish 2 Decision support. Development of expert the processes and resources required for successful system/knowledge-based systems to help implementation. The concept of segmentation, management to select and manage the portfolio once adjusted to reflect the impact of the informa- of segments seems vital. Such systems would tion revolution and the globalization of business, is ideally be incorporated in an effective DSS.The sound and valid. It is the practice of segmentation key to this is the development of a set of rules that is fraught with problems. These problems are summarizing our current understanding of solvable but the solutions require revision and market segmentation.These rules can reflect the alteration to most of the current approaches used to empirical generalizations in this area and can be segment markets. If we have the conviction and aided by appropriate meta-analyses. In addition, courage to re-examine the traditional segmentation more thought needs to be given to the concept and approaches, we shall be able to signifi- psychological and other impediments to use and cantly increase the value of our segmentation implementation (see e.g. Hoch and Schkade, efforts in creating value to our customers and other 1996). stakeholders. Ch11-H8566.qxd 8/8/07 2:04 PM Page 242

242 The Marketing Book

Christensen, C., Cook, S. and Hall, T. (2005) Mar- References keting malpractice: the cause and cure, Harvard Business Review, HBR Reprint 2386. Allenby, G. M., Arora, N. and Ginter, J. L. (1998) Cooper, L. G. and Giuffrida, G. (2000) Turning On the heterogeneity of demand, Journal of datamining into a management science tool, Marketing Research, 35(3), 384–389. Management Science, 47(2), 249–264. Allenby, G. M. and Rossi, P. E. (1999) Marketing Czinkota, M. R., Ronkainen, I. A. and Donath, B. models of consumer heterogeneity, Journal of (2004) Mastering Global Markets: Strategies for Econometrics, 89, 57–78. Today’s Trade Globalist, South-Western, Mason, Anderson, C. (2006) The Long Tail: Why the Future OH. of Business is Selling Less of More, Hyperion. Dahan, E. and Hauser, J. R. (2002) The virtual cus- Arabie, P., Douglas, C. J., DeSarbo, W. S. and tomer, Journal of Product Innovation Manage- Wind, J. (1981) Overlapping clustering: a new ment, 19(5), 332–347. method for product positioning, Journal of Mar- Dahan, E. and Srinivasan, V. (2000) The predictive keting Research, 18, 310–317. power of internet-based product concept test- Aribarg, A., Arora, N. and Onur Bodur, H. (2002) ing using visual depiction and animation, Jour- Understanding the role of preference revision nal of Product Innovation Management, 17, and concession in group decisions, Journal of 99–109. Marketing Research, 39(3), 336–349. Eisenbeis, R. A. (1977) Pitfalls in the application of Arora, N. (2006) Estimating joint preference: a discriminant analysis in business, finance, and subsampling approach, International Journal of economics, Journal of Finance, 33(3), 875–900. Research in Marketing, 23(4), 409–421. Elrod, T. and Keane, M. P. (1995) A factor-analytic Arora, N. and Allenby, G. M. (1999) Measuring probit model for representing the market struc- influence of individual preference structures in ture in panel data, Journal of Marketing Research, group decision making, Journal of Marketing 32(1), 1–16. Research, 36(4), 476–487. Frank, R., Massy, W. and Wind, J. (1972) Market Seg- Blattberg, R., Glazer, R. and Little, J. (1994) The mentation, Prentice Hall, Englewood Cliffs, NJ. Marketing Information Revolution, Harvard Busi- Garreau, J. (1982) The Nine Nations of North Amer- ness School Press, Harvard. ica, Houghton Mifflin, Boston. Bell, D. R., Ho, T.-H. and Tang, C. S. (1998) Deter- Gensch, D., Aversa, N. and Moore, S. (1990) A mining where to shop: fixed and variable costs choice modeling market information system of shopping, Journal of Marketing Research, 35(3), that enabled ABB Electric to expand its market 352–369. share, Interfaces, 20, January–February, 6–25. Bell, D. R. and Song, S. (2007) Neighborhood Gilbride, T. J. and Allenby, G. M. (2004) A choice effects and trial on the internet: evidence from model with conjunctive, disjunctive, and com- online grocery retailing, Quantitative Marketing pensatory decision rules, Marketing Science, and Economics, forthcoming. 23(3), 391–406. Bucklin, R. E. and Gupta, S. (1992) Brand choice, Gilbride, T. J. and Allenby, G. M. (2006) Estimating purchase incidence, and segmentation: an inte- heterogeneous EBA and economic screening grated modeling approach, Journal of Marketing rule choice models, Marketing Science, 25(5), Research, XXIX, May, 201–215. 494–511. Camm, J. D., Cochran, J. J., Curry, D. J. and Green, P. E. and DeSarbo, W. S. (1979) Componen- Kannan, S. (2006) Conjoint optimization: an tial marketing in the analysis of consumer exact branch-and-bound algorithm for the trade-offs, Journal of Marketing, 43, 83–91. share-of-choice problem, Management Science, Green, P. E. and Krieger, A. M. (1985) Models and 52(3), 435–447. heuristics for product line selection, Marketing Chaturvedi, A., Douglas, C. J. and Green, P. E. Science, 4, Winter, 1–19. (1994) Market Segmentation via Overlapping K- Green, P. E. and Krieger, A. M. (1991) Segmenting centroids Clustering, Working Paper, Wharton marketing with conjoint analysis, Journal of School, University of Pennsylvania, Pennsylva- Marketing, 55, October, 20–31. nia, May, 1–21. Green, P. E. and Krieger, A. M. (1994) An Evaluation Choffray, J.-M. and Lilien, G. L. (1978) A new of Alternative Approaches to Cluster-Based Market approach to industrial market segmentation, Segmentation, Working Paper, Wharton School, Sloan Management Review, 19(3), Spring, 17–30. University of Pennsylvania, Pennsylvania. Ch11-H8566.qxd 8/8/07 2:04 PM Page 243

Market segmentation 243

Green, P. E., Krieger, A. M. and Wind, Y. (2003) Kotler, P. and Keller, K. L. (2005) Marketing Buyer Choice Simulators, Optimizers and Management, Prentice Hall, Englewood Cliffs, Dynamic Models, in Wind, J. and Green, P. E. NJ. (eds), Market Research and Modeling: A Tribute to Levin, N. and Zahavi, J. (2001) Predictive model- Paul E. Green, Kluwer Academic Publishers, ing using segmentation, Journal of Interactive Norwell, MA. Marketing, 15(2), Spring, 2–22. Grover, R. and Srinivasan, V. (1987) A simultan- Lewinson, L. (1994) Data mining: tapping into the eous approach to market segmentation and mar- mother code, Database Programming and Design, ket structuring, Journal of Marketing Research, 24, February, 50–56. May, 139–153. Lilien, G. and Rangaswamy, A. (2002) Marketing Grover, R. and Srinivasan, V. (1992) Evaluating Engineering: Computer-Assisted Marketing Analy- the multiple effects of retail promotions on sis and Planning, Trafford, Victoria, BC. brand loyal and brand switching segments, Little, J. D. C. (1979) Decision support systems for Journal of Marketing Research, 29(1), 76–89. marketing managers, Journal of Marketing, Gupta, S. and Lehmann, D. R. (2003) Customers as 43(3), 9–26. assets, Journal of Interactive Marketing, 17(1), 9–24. Lodish, L. (2001) Building marketing models that Harris, R., Sleight, P. and Webber, R. (2005) Geode- make money, Interfaces, 31(3), 45–55. mographics, GIS and Neighborhood Targeting, Moe, W. W. and Fader, P. S. (2001) Modeling Wiley, Hoboken, NJ. hedonic portfolio products: a joint segmentation Helsen, K., Jedidi, K. and DeSarbo, W. S. (1993) A analysis, Journal of Marketing Research, 38(3), new approach to country segmentation utiliz- 376–385. ing multinational diffusion patterns, Journal of Moriarty, R. T. and Reibstein, D. J. (1986) Benefit Marketing, 57, October, 60–71. segmentation in industrial markets, Journal of Hill, S., Provost, F. and Volinksy, C. (2006) Business Research, 14(6), 463–486. Network-based marketing: identifying likely Peppers, D. and Rogers, M. (1993) The One to One adopters via consumer networks, Statistical Sci- Future, Doubleday, New York. ence, 22(2), 256–276. Peterson, R. A. (1974) Market structuring by Hoch, S. J. and Schkade, D. A. (1996) A psycho- sequential cluster analysis, Journal of Business logical approach to decision support systems, Research, 2, July, 249–264. Management Science, 42(1), 51–64. Politis, D., Romano, J. P. and Wolf, M. (1999) Sub- Iyengar, R., Jedidi, K. and Kohli, R. (2007) A con- sampling, Springer Series in Statistics, Springer- joint approach to multi-part pricing, Journal of Verlag, New York, NY. Marketing Research, forthcoming. Rangaswamy, A. and Wind, Y. (1994) Don’t Walk Johnson, B. and Lilien, G. L. (1994) A Framework In, Just Log In! Electronic Commerce on the Infor- and Procedure for Assessing Market Segment mation Highway, Working Paper, Wharton Change, Penn State ISBM Working Paper, Penn School, February. State University, University Park, PA. Rayward-Smith, V. (2002) Fuzzy cluster analysis: Kamakura, W. and Russell, G. J. (1989) A prob- methods for classification, data analysis and abilistic choice model for market segmentation image recognition, The Journal of the Operational and elasticity structure, Journal of Marketing Research Society, 51(6), 769–780. Research, 26(4), 379–390. Reutterer, T., Mild, A., Natter, M. and Taudes, A. Kamakura, W. and Wedel, M. (2004) An empirical (2006) A dynamic segmentation approach for Bayes procedure for improving individual-level targeting and customizing direct marketing estimates and predictions from finite mixtures campaigns, Journal of Interactive Marketing, of multinomial logit models, Journal of Business 20(3/4), 43–52. and Economic Statistics, 22(1), 121–125. Rossi, P. E. and Allenby, G. M. (2003) Bayesian stat- Kamakura, W., Kim, B.-D. and Lee, J. (1996) Mod- istics and marketing, Marketing Science, 22(3), eling preference and structural heterogeneity 304–328. in consumer choice, Marketing Science, 15(2), Rossi, P. E., McCulloch, R. E. and Allenby, G. M. 152–172. (1996) The value of purchase history data in Kamakura, W., Kossar, B. S. and Wedel, M. (2004) target marketing, Marketing Science, 15(3), Identifying innovators for the cross-selling 321–340. of new products, Management Science, 50(8), Saaty, T. L. (1992) Decision Marketing for Leaders, 1120–1133. RWS Pittsburgh Publications, Pittsburgh. Ch11-H8566.qxd 8/8/07 2:04 PM Page 244

244 The Marketing Book

Sandor, Z. and Wedel, M. (2005) Heterogenous segmentation, Journal of Marketing Research, conjoint designs, Journal of Marketing Research, XXVIII, November, 385–396. 42(2), 210–220. Weiss, M. (1994) Latitudes of Altitudes: An Atlas Schweidel, D. A., Bradlow, E. T. and Williams, P. of American Tastes, Trends, Politics and Passions, (2006) A feature-based approach to assessing Little Brown and Company, Boston. advertising similarity, Journal of Marketing Wilson, E. J., Lilien, G. L. and Wilson, D. T. (1991) Research, 43(2), 237–243. Developing and testing a contingency para- Sleight, P. (2004) Targeting Customers, 3rd edn, digm of group choice in organizational buying, World Advertising Research Council (WARC), Journal of Marketing Research, 28, November, London. 452–466. Srinivasan, V., Lovejoy, W. S. and Beach, D. (1997) Wind, Y. (1978) Issue and advances in segmenta- Integrated product design for marketability and tion research, Journal of Marketing Research, XV, manufacturing, Journal of Marketing Research, August, 317–337. 34(1), 154–163. Wind, Y. (2007, forthcoming) The power of adap- Steenkamp, J.-B. E. M. and Ter Hofstede, F. (2002) tive experimentation, Marketing Research. International market segmentation: issues and Wind, Y. and Cardozo, R. N. (1974) Industrial perspectives, International Journal of Marketing marketing segmentation, Industrial Marketing Research, 19, 185–213. Management, March, 153–165. Ter Hofstede, F., and Park, K. (2007) On the Spatial Wind, Y. and Robertson, T. (1983) Marketing strat- Structure of Consumer Needs: Cross-National Evi- egy: new directions for theory and research, dence from the European Union, Working Paper, Journal of Marketing, Spring, 12–25. Coombs School of Business, UT Austin. Wind, Y. and Saaty, T. L. (1980) Marketing appli- Ter Hofstede, F., Wedel, M. and Steenkamp, J.-B. cations of the analytic hierarchy process, Man- E. M. (2002) Identifying spatial segments in agement Science, 26, July, 641–658. international markets, Marketing Science, 21(2), Wind, Y., Crook, C. and Gunther, R. (2004) The 160–177. Power of Impossible Thinking: Transform the Busi- Toubia, O., Hauser, J. R. and Garcia, R. (2007) ness of Your Life and the Life of Your Business, Probabilistic polyhedral methods for adaptive Wharton School Publishing, Upper Saddle choice-based conjoint analysis: theory and appli- River, NJ. cation, Marketing Science, forthcoming. Yankelovich, D. and Meer, D. (2006) Rediscover- Wedel, M. and Steenkamp, J.-B. E. M. (1991) A ing market segmentation, Harvard Business clusterwise regression method for simultane- Review, 84(2), 122–132. ous fuzzy market structuring and benefit