International market segmentation: issues and perspectives ?· International market segmentation: issues…

  • Published on

  • View

  • Download


International market segmentation: issues and perspectivesJan-Benedict E.M. Steenkamp a,*, Frenkel Ter Hofstede baTilburg University, P.O. Box 90153, 5000 LE Tilburg, The NetherlandsbCarnegie Mellon University, Graduate School of Industrial Administration, 5000 Forbes Avenue, Pittsburgh, PA 15213, USAAbstractWith the increasing globalization of the business world, international segmentation becomes an ever more important conceptin marketing. The globalization forces now at work push many companies to extend or reorganize their marketing strategiesacross borders and target international segments of consumers. It is the purpose of this paper to review the international marketsegmentation literature and to identify its future prospects and threats. We critically assess the current status of internationalmarket segmentation research and provide a systematic overview of 25 previous empirical studies with respect to the samplesused for segmentation, segmentation bases and methods, geographic configuration of segments, and validation efforts. Wediscuss a number of conceptual and methodological issues that deserve more attention if international market segmentation is tofulfill its high potential. The conceptual issues include construct equivalence of the segmentation basis used, level of aggregationin the segmentation process, and choice of the segmentation basis. The methodological issues include measure equivalence andsample equivalence of the segmentation basis, segmentation methods employed, and whether national sample sizes should beproportional to population sizes. We describe a case study to illustrate and integrate the various issues and conclude withsuggestions for future research to stimulate further advances in the area.D 2002 Elsevier Science B.V. All rights reserved.Keywords: International market segmentation; International marketing; Segmentation1. IntroductionInternational market segmentation has become animportant issue in developing, positioning, and sellingproducts across national borders. It helps companiesto target potential customers at the international-seg-ment level and to obtain an appropriate positioningacross borders. A key challenge for companies is toeffectively deal with the structure of heterogeneity inconsumer needs and wants across borders and totarget segments of consumers in different countries.These segments reflect geographic groupings orgroups of individuals and consist of potential con-sumers who are likely to exhibit similar responses tomarketing efforts.A natural form of international segmentation is toadopt a multi-domestic strategy where each countryrepresents a separate segment (Jeannet & Hennessey,1998). A multi-domestic strategy amounts to selectionof countries on the basis of their local advantages.Traditionally, multinational companies implementedsuch multi-domestic strategies by tailoring nationalbrands to the needs shared by groups of consumers0167-8116/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.PII: S0167 -8116 (02 )00076 -9* Corresponding author. Tel.: +31-13-4662916; fax: +31-13-4662875.E-mail address: (J.-B.E.M. Steenkamp) J. of Research in Marketing 19 (2002) 185213in the same country. In such an approach no coordi-nation is required between countries, products areproduced locally and are tailored to satisfy local needs.Distinct advertising, distribution, and pricing strategiesare developed for targeting consumers in each country,and competition is managed at a national level. Com-petitive moves are conducted on a country-to-countrybasis and do not take the developments in othercountries into account. Thus, in segmenting theirmarkets, firms operating according to a multi-domesticapproach can suffice with the standard segmentationtechniques that are developed for domestic markets(Jeannet & Hennessey, 1998).International segmentation becomes a particularlychallenging issue when companies adopt a global orpan-regional strategy, that is, a strategy integratedacross national borders. In many industries, nationalborders are becoming less and less important as anorganizing principle for international activities, render-ing multi-domestic strategies less relevant (Yip, 1995).Developments accelerating this trend include regionalunification, shifts to open economies, global invest-ment, manufacturing, and production strategies,expansion of world travel, rapid increase in education,literacy levels, and urbanization among developingcountries, convergence of purchasing power, life-styles and tastes, advances in information and commu-nication technologies, the emergence of global media,and the increasing flow of information, labor, money,and technology across borders (Gielens & Dekimpe,2001; Hassan & Katsanis, 1994; Hassan & Kaynak,1994; Parker & Tavassoli, 2000; Yip, 1995). Manyglobal companies such as Coca-Cola, McDonalds,Sony, British Airways, Ikea, Toyota, and Levi-Strausshave successfully integrated their international strat-egies. The forces that are now at work drive manycompanies to extend their operations abroad and targetinternational market segments. By globalizing theirstrategies, such companies benefit from several advan-tages, including cost reductions through economies ofscale, improved quality of products, and increasedbargaining and competitive power (Levitt, 1983; Yip,1995).Still, companies cannot serve the entire heteroge-neous population of (a region of) the world with fullystandardized marketing strategies. Many companiesrecognize that groups of consumers in different coun-tries often have more in common with one another thanwith other consumers in the same country. Hence, theychoose to serve segments that transcend nationalborders (Hassan & Katsanis, 1994). International seg-mentation aids the firm in structuring the heterogeneitythat exists among consumers and nations and helps toidentify segments that can be targeted in an effectiveand efficient way. As argued by Walters (1997, pp.165166):When significant heterogeneity characterizes theinternational market context, tools are neededwhich can assist in the identification of under-lying patterns of similarity which can provide aplatform for global integration at the strategic andoperational levels. The segmentation constructoffers great promise in this respect. . .. Segmenta-tion is therefore particularly important in enter-prises that wish to develop and implementsuccessful global marketing strategies.International segmentation offers a solution to thestandardization versus adaptation debate in that itcreates the conceptual framework for offering productsand/or marketing programs that are standardizedacross countries by targeting the same consumer seg-ment(s) in different countries (Verhage, Dahringer, &Cundiff, 1989). When using a similar marketing strat-egy in multiple countries, economies of scale will leadto a reduction in the average costs of production,advertising, and distribution. If, at the same time,consumers in the targeted segments share the sameneeds, such strategies can also be highly effective (Yip,1995). Hence, international segmentation combinesthe benefits of standardization (e.g., lower costs, betterquality) with the benefits of adaptation (e.g., close toneeds of consumers).Despite the obvious importance of internationalmarket segmentation for marketing as a discipline ingeneral and international marketing in particular, it hasreceived relatively little attention in the literature. In areview of about 900 articles on international marketing,it was found that just over 1% (11 papers) dealt directlywith international market segmentation (Aulakh &Kotabe, 1993). A similar observation applies to inter-national versus domestic market segmentation: Seg-mentation is a central issue in domestic marketingstrategy. Yet, in international markets, it has receivedlittle attention (Douglas & Craig, 1992, p. 312). OneJ.-B.E.M. Steenkamp, F. Ter Hofstede / Intern. J. of Research in Marketing 19 (2002) 185213186exception is the review by Walters (1997), whichfocuses especially on the process and internationalbusiness aspects of international segmentation. Weexpand upon this work by providing an in-depth treat-ment of the key conceptual and methodological issuesone should address when doing international segmen-tation research. Our point of departure is a systematicoverview of previous empirical international segmen-tation studies. These studies are subsequently related tothe various issues raised in international market seg-mentation. These conceptual and methodologicalissues deserve more attention if international marketsegmentation is to fulfill its potential. We bring theseissues together and apply them in an illustrative casestudy involving an international segmentation chal-lenge faced by one of the largest global consumerpackaged good companies. We conclude with sugges-tions for future research on international market seg-mentation.2. Previous empirical research in internationalsegmentationTable 1 provides an overview of the key features of25 international market segmentation studies. For eachstudy, we provide information on the sample, segmen-tation basis,1 segmentation method, results, and vali-dation, and briefly comment on some notable featuresof the study. In Fig. 1 the same studies are classifiedaccording to (1) their geographic coverage (whethernon-Triad countries were included in the study ornot),2 (2) the level of the segmentation basis used(country, region, or consumer), and (3) the nature ofthe study (exploratory versus model-based). AlthoughTable 1 and Fig. 1 are more or less self-explanatory,several comments are in order.First, whereas the total number of studies on inter-national market segmentation has been limited up to1990, it is encouraging to see that the interest ininternational segmentation increases over time, asevidenced by the relative large number of studies thatwere published after 1990 (14 out of 25 studies).Second, international market segmentation researchis truly international in that many different countriesfrom all (inhabited) continents are included. Althoughthere is a bias toward Triad countries (see Fig. 1), it isencouraging to see that research has also been con-ducted in other parts of the world. Hofstedes (1980)seminal study may have served as exemplar.Third, the studies have used a wide variety ofsegmentation bases and segmentation methods. Somestudies used information on countries (or regionswithin countries), aggregated across consumers tothe country (or region) level or information pertainingto the countries (or regions) themselves (e.g., climate,legal regime). Other studies used disaggregate, indi-vidual-specific information of consumers. The coun-try-level segmentation bases typically included acombination of economic, political, geographic anddemographic information (e.g., Helsen, Jedidi, &DeSarbo, 1993; Huszagh, Fox, & Day, 1986) orcultural variables (e.g., Hofstede, 1980; Sirota &Greenwood, 1971; Steenkamp, 2001). The set ofindividual-level segmentation bases is more heteroge-neous, ranging from domain-specific characteristicssuch as attribute evaluations (Moskowitz & Rabino,1994), attitudes (Verhage et al., 1989), and risk andbrand loyalty ratings (Yavas, Verhage, & Green, 1992)to intermediate constructs such as means-end chains(Ter Hofstede, Steenkamp, & Wedel, 1999) to generalcharacteristics such as values (Kamakura, Novak,Steenkamp, & Verhallen, 1993). The diversity ofsegmentation bases attests to the versatility of interna-tional market segmentation, but there is an overrepre-sentation of studies using characteristics of countries.Only eight studies used responses from individualconsumers as compared to fifteen using country-levelbases. Information on countries often is acquiredrelatively easily through published secondary datasources (e.g., publications of Euromonitor, the UnitedNations, and the World Bank), but the results of suchcountry clusters do not always provide relevant infor-mation for managerial decisions (Helsen et al., 1993;Nachum, 1994). Presumably, the lack of availablecross-national consumer data and the difficulty andcost involved in its collection, relative to the avail-1 The segmentation basis is the set of criteria used to groupconsumers into international segments.2 Following Ohmae (1985), a distinction is made between Triadand non-Triad countries. Collectively, the Triad regions produce theoverwhelming share of the worlds GDP, and participation in theTriad regions is generally recommended (Jeannet & Hennessey,1998). This renders market segmentation focusing on these regionsespecially relevant.J.-B.E.M. Steenkamp, F. Ter Hofstede / Intern. J. of Research in Marketing 19 (2002) 185213 187Table 1Overview of previous empirical international market segmentation studiesStudy Sample Segmentation basis Method Results Validation Comments1 Askegaardand Madsen(1998)20,000 respondentsfrom 79 regionsfrom 15European countriesaverage responseper region to 138food-related items(i) factoranalysis onregion-levelresponses(ii) clusteranalysison regionscoreson 41 factorsextracted12 geographicallycontiguous segmentsfollowing languagebordersseveral factoranalyticand clustermethodsyielded largelythe same resultsdominance of country-specific segmentsmay be partially dueto translation bias2 Boote (1983) 899 respondentsfrom UK,Germany, France29 psychographicstatementsQ factoranalysis onindividualresponses4 segmentsof which 2have substantialmembershipin all 3 countriesno no examination ofcross-nationaldifferences in responsetendencies3 Dawar andParker (1994)35 countriescoveringall continents% of labor forceengaged in retailsectorcluster analysis 2 segments,high versuslow engagementno no difference foundbetween clusters onmeasures of interest-likelihood of usingeach of four quality cues4 Dayet al. (1988)96 countriescoveringall continents18 countrycharacteristicsreflecting economicdevelopment(i) factoranalysis oncountrycharacteristics(ii) clusteranalysis oncountryscores onthree factorsextracted6 geographicallydispersed segmentsno segments varied muchin number of countries;large differences withthe (comparable)Sethi (1971) studyJ.-B.E.M.Steenkamp,F.TerHofstede/Intern.J.ofResearchinMarketing19(2002)1852131885 Helsenet al. (1993)12 Triad countries 23 countrycharacteristicsreflectingeconomicdevelopment(i) factoranalysis oncountrycharacteristics(ii) cluster analysison country scores onfive factors retained2 and 3segmentsolutionsexamined;both solutionscontained alarge segment ofmost WestEuropeancountrieslittle similarity withproduct-specificsegmentation(see below)similar to Sethi (1971),USA was outlier(in three-segment solution)6 Helsenet al. (1993)12 Triad countries annual unit salesdata per countryfor TV, VCR, andCD-playersfor 14 yearslatent class estimationof the Bass modelper product3 segments (TV, VCR)and two segments(CD-player),geographicallydispersedlittle similarity withsociodemographicsegmentation(see above)large differences among the3 segmentation solutionsindicate limitations ofcountry-based segmentationusing product-specific bases7 Hofstede(1976)315 managers from14 countries coveringfour continents12 scales forpersonal andinterpersonalvaluesQ factor analysis oncountry averagescores on the scales5 segments, ofwhich two werecontiguoussubstantial agreementin within-segmentvalue profiles withstudent dataFrench (German)-speakingSwitzerland was assignedto Latin (Germanic)segment, providingevidence for within-countrycultural value heterogeneity8 Hofstede(1980)88,000 employeesfrom 40 countriescovering allcontinentsvarying sets ofitems related towork goals(i) theoretical reasoningand factor analysison country averageitem scores(ii) cluster analysis oncountry scores on thefour dimensionsretained followed byjudgmental regroupings8 segments, most ofwhich tended tobe geographicallycloseresults partiallyreplicated with SSAHofstedes four dimensionsprovide the bestvalidated and most appliedoperationalization ofnational culture(continued on next page)J.-B.E.M.Steenkamp,F.TerHofstede/Intern.J.ofResearchinMarketing19(2002)185213189Table 1 (continued)Study Sample Segmentation basis Method Results Validation Comments9 Huszaghet al. (1986)21 Triad countries 9 countrycharacteristicsreflecting economicdevelopmentcluster analysis oncountry scores5 segments of whichone was geographicallycontiguouslargely same resultswere obtained fordifferent clusteringalgorithms and forslightly varying setsof characteristicshigh variation in productacceptance rates across 27product categories for 5European countriesassigned to samesegments sheds doubton usefulness ofsegmentation basis10 Kale (1995) 17 WesternEuropeancountriesHofstedes (1980)national-culturaldimensionscluster analysis oncountry ratings on theHofstede dimensions3 segments, none ofthem completelycontiguousno only one segment thesame as in Hofstede (1980);results also very differentfrom other related studies(Ronen & Shenkar, 1985)11 Kamakuraet al. (1993)1573 consumersfrom Italy,West Germany,UKRokeach instrumentaland terminal valuesclusterwise rank logiton individual-levelrankings of the top-nine instrumental andterminal values5 segments, allshowing a strongnationalorientationincluding segmentinformation over countryinformation leads tosignificant improvementin R2 for 40 out of59 psychographicand benefit itemsincomplete rankings reduceddiscriminability of segments12 Kumaret al. (1998)14 WesternEuropeancountriesannual unit sales dataper country for 5consumer durablesfor 1970s1990cluster analysis on theBass coefficients ofinnovation and imitationper product3 segments per product;segment compositionnot consistent acrosscountriesno correspondencewith sociodemographicsegmentation ofHelsen et al. (1993)cluster solutions indicatethat countries tend to groupbased on time ofintroduction, geographicalproximity, and cultural/economic similarity13 Kumaret al. (1994)11 WesternEuropeancountriesmarket potential ofeach country on 6criteria in 8different machineryindustriesinteractive multicriteriaapproach for solvingproblems of multiobjectivedecision modelsusing Augmented WeightedTchebycheff proceduredifferent segments ofcountries industries arefound dependent ontradeoffs between criteriamulticriteria approachsuperior to personalevaluation by managersthe only study incorporatinginternational marketsegmentation in decisionsupport systems14 Lee (1990) 70 countriescoveringall continents4 countrycharacteristicsreflecting economicdevelopment(i) stepwise regression ofnational innovativenesson 10 countrycharacteristics(ii) cluster analysis oncountry scores on foursignificant characteristics5 segments eachcontaining countriesfrom at least 2continentsno segments interpreted usingRogers (1995) schemebut no scores on nationalinnovativeness providedJ.-B.E.M.Steenkamp,F.TerHofstede/Intern.J.ofResearchinMarketing19(2002)185213190(continued on next page)15 Moskowitzand Rabino(1994)808 consumersfrom 4 (unidentified)countries10 sensory attributesof flavored soda(i) factor analysis onrespondents idealpoints on sensory attributes(ii) cluster analysis onrespondents factor scores3 cross-nationalsegmentsno cross-national segmentsgenerate larger variation inoptimal productformulation thancountries16 Ronen andKraut (1977)4000 techniciansfrom 14 mostlyEuropean countries22 items related towork goalssmallest space analysis(SSA) on rank ordercorrelations across theaverage scores on the 22items for each pair ofcountries and visualclustering on SSA locations4 segments, of which2 are contiguousno choice of segments basedon a priori notions and notstrongly supported byactual results17 Sethi (1971) 91 countriescoveringall continents29 countrycharacteristicsreflecting economicdevelopment(i) factor analysis oncountry characteristics(ii) cluster analysis oncountry scores onfour factors extracted7 geographicallydispersed segmentsno segments varied much innumber of countries;USA was outlier18 Sirota andGreenwood(1971)about 13,000employeesfrom 25 countriescoveringall continents14 items related towork goalsQ factor analysison country averagescores on the items6 segments, of whichfour are contiguousRonen and Kraut(1977) reanalyzed thedata with SSA andvisual clustering andmany differenceswere foundsegment of 6independents seems restcategory indicating thatabout 25% ofcountries could not beproperly assigned19 Steenkamp(2001)24 countriescovering5 continentsHofstedes (1980)4 and Schwartzs(1994) 7 nationalcultural dimensions(i) factor analysison country scores on11 dimensions(ii) two-stage clusteranalysis on countryscores on 4 factors extracted7 segments, 4 of whichgeographicallylargely contiguousno only study using Schwartzstheory-derived frameworkof national culture20 Ter Hofstedeet al. (1999)2961 consumersfrom 11 EUcountriesmeansend chaindata on yogurtmixture modelincorporating within- andbetween-countryheterogeneity in responsebehavior using countrymembership as concomitantvariables4 cross-nationalsegments, includingone pan-Europeansegmentassessment ofpredictive validity;comparison withK-means clusteringhigh predictive validity;model outperforms clusteranalysis; significantdifferences in responsetendencies found withinand between countriesJ.-B.E.M.Steenkamp,F.TerHofstede/Intern.J.ofResearchinMarketing19(2002)185213191Study Sample Segmentation basis Method Results Validation Comments21 Ter Hofstedeet al. (2002)1966 consumersfrom 120 regionsin 7 EU countriesimportances of 6image attributesfor outlets sellingmeat withineach regionconstrained hierarchicalBayes model based onsimilarity in inferred imageimportances of consumersin different regionsidentifying contiguousgeographic segments5 contiguous segmentsof which 3 arecross-national and 2largely country-specificcomparison withunrestrictedsegmentation andcountry segmentationindicated strongperformance of modelmodel accommodates abroad set of constraints onsegments, which allowsmanagerial objectives tobe included in thesegmentation process22 Vandermerweand LHuillier(1989)173 regions from 18European countries5 geographic andsociodemographicregion-levelcharacteristicscluster analysis onregion scores6 geographicallycontiguous segmentstranscendingnational bordersno choice of region-levelvariables biases results tocontiguous segments;variables range fromnominal to ratio levelcreating problems incluster analysis23 Verhageet al. (1989)571 consumers fromFrance, Netherlands,Norway, USA14 attitudes towardenergy conservationcluster analysis onindividual-levelratings3 cross-nationalsegmentsQ factor analysis andcluster analysisindicated same numberof clusters; significantdifferences betweensegments for 9 out of18 conservation behaviorsscreen test indicates2 segments: segments2 and 3 not wellseparated on attitudes24 Wedel et al.(1998)3900 consumersfrom 6 EUcountriesrank order of 9LOV valuesmixture model whichaccounts for differentsampling designs,including stratificationby country5 cross-nationalsegments including2 pan-European andone country-specificsegmentcomparison withmixture model whichignores complexsampling designaccounting for complexsampling design leads todifferent and betterinterpretable results25 Yavas et al.(1992)781 consumers from6 countries coveringthree continentsrisk and brand loyaltyratings for bathsoap and toothpastecluster analysis onindividual-level ratings4 cross-nationalsegmentsno conceptual relationbetween the productsused for segmentationnot clearTable 1 (continued)J.-B.E.M.Steenkamp,F.TerHofstede/Intern.J.ofResearchinMarketing19(2002)185213192ability of information about countries, has discouragedresearchers to study international consumer segments.Fourth, it should be noted that most studies are of arather exploratory nature, especially the studies thatsegmented countries. As shown in Fig. 1, few studiestest specific hypotheses or base their work on aspecific conceptual model (e.g., Helsen et al., 1993;Kumar, Ganesh, & Echambadi, 1998; Ter Hofstede etal., 1999). The exploratory nature of the studies limitsthe generalizability of specific findings. In Section4.3, we will elaborate on methodological develop-ments relevant for international segmentation.Fifth, the studies involving responses from indi-vidual consumers have been conducted mainly inTriad countries. One exception is the study of Yavaset al. (1992), which also included consumers fromThailand and Saudi Arabia. A reason for the focus onTriad countries is their economic importance. Inaddition, the market research agency infrastructurein Triad countries is better developed and respondentsmay be more familiar with the instruments (e.g., ratingscales) often used in such studies. Advanced market-ing research is feasible, however, among less favoredpopulations provided appropriate safeguards are takensuch as adaptation of the measurement instrument(Steenkamp & Burgess, 2002).Sixth, despite the growing importance of marketingdecision support systems (Van Bruggen & Wierenga,2000), and the important role market segmentation canplay in such systems, only one study has integratedFig. 1. Classification of international segmentation studies according to geographic coverage, level of aggregation of the segmentation basis, andnature of the study.J.-B.E.M. Steenkamp, F. Ter Hofstede / Intern. J. of Research in Marketing 19 (2002) 185213 193international market segmentation in a decision sup-port system (Kumar, Stam, & Joachimsthaler, 1994).They combined information on product-country mar-kets with decision-makers objectives, constraints, andstrategies, to identify the most attractive product-country segments through an interactive process.Seventh, little attention is given to the validation ofthe solution. Eleven studies conducted no validationeffort while the validation in most other studies waslargely limited to an informal comparison of resultsobtained by different estimation methods (e.g., Aske-gaard & Madsen, 1998; Huszagh et al., 1986). One ofthe exceptions is Kamakura et al. (1993). They explic-itly tested the statistical significance of the increase inexplained variance in 59 psychographic and benefititems when international segment information wasadded to country information. If the international seg-ments add insights over and above country member-Fig. 2. Issues in International Market Segmentation.J.-B.E.M. Steenkamp, F. Ter Hofstede / Intern. J. of Research in Marketing 19 (2002) 185213194shipwhich may be considered as an a priori geo-graphic segmentation schemethe increase in R2should be significant. This was the case for 40 out of59 variables, which is far better than chance ( p