This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Journal of Retailing 86 (4, 2010) 400–413

Exploring the post-termination stage of consumer–brand relationships: Anempirical investigation of the premium car market

Gaby Odekerken-Schröder a,∗, Thorsten Hennig-Thurau b,c,1, Anne Berit Knaevelsrud d,2

a Maastricht University, School of Business and Economics, Department of Marketing & Supply Chain Management, P.O. Box 616, 6200 MD Maastricht,The Netherlands

b University of Muenster, Marketing Center Muenster, Am Stadtgraben 13-15, 48143 Muenster, Germanyc Cass Business School, Department of Management, 106 Bunhill Row, London EC1Y 8TZ, UK

d Bauhaus-University of Weimar, Faculty of Media, Helmholtzstraße 15, 99425 Weimar, Germany

Abstract

This research extends customer lifecycle models to include a post-termination stage that bridges the dissolution stage of a consumer–brandrelationship with a potential recovery stage. Drawing from 43 depth interviews with former customers of a car brand, this study relies on groundedtheory and triangulation to explore consumer responses in the post-termination stage and finds evidence for both negative and positive customerresponses after dissolution. The authors combine qualitative techniques with categorical principal component analysis (CATPCA) to explorethe dimensionality of post-termination responses and extract relationship-related and termination-related response dimensions. They use thesedimensions to identify four distinct post-termination customer clusters, which differ systematically with the customer’s termination reasons,as demonstrated through a series of exact logistic regressions. In addition to providing evidence of a post-termination stage, this study offersimplications for customer relationship management.© 2010 New York University. Published by Elsevier Inc. All rights reserved.

Keywords: Retail patronage; Post relationships; Relationship termination; Grounded theory; Regain management; Car industry; CATPCA; Cluster analysis; Exactlogistic regression

The study of patronage behavior is critical to retailingand relationship marketing, yet knowledge about consumers’responses after they terminate a relationship with a companyor brand remains rare (e.g., Coulter and Ligas 2000; Grewalet al. 2003; Pan and Zinkhan 2006). Since Dwyer, Schurr, andOh (1987) included a dissolution stage in their seminal workon buyer–seller relationships, literature on service failure andrecovery has pointed at critical behaviors that lead to dissolu-tion (Brady et al. 2008; Keaveney 1995; Mittal, Huppertz, andKhare 2008), and some other studies have investigated relation-ship managers’ “win-back” strategies (e.g., Stauss and Friege1999; Thomas, Blattberg, and Fox 2004; Tokman, Davis, andLemon 2007), as well as consumers’ perceptions of recoveryefforts (Mattilla and Patterson 2004). A systematic analysis of

∗ Corresponding author. Tel.: +31 43 388 3618; fax: +31 43 388 4918.E-mail addresses: g.odekerken@maastrichtuniversity.nl

(G. Odekerken-Schröder), thorsten@hennig-thurau.de (T. Hennig-Thurau),anne.knaevelsrud@uni-weimar.de (A.B. Knaevelsrud).

1 Tel.: +49 251 83 22808; fax: +49 251 83 22903.2 Tel.: +49 3643 58 38 22; fax: +49 3643 58 37 91.

consumer behavior after relationship dissolution, however, ismissing.

A potential reason for that is that no consumer responses areexpected to exist after dissolution. Anecdotal evidence indicatesthat managers believe lost customers are no longer emotionallyattached to a former brand; according to the Head of CustomerOperations at Hutchinsons 3G, Austria Ltd., “A customer whodecides to leave a company has lost his emotional attachment”(Baumgartner 2008, p. 37). However, using social relationshipsas an analogy, we argue that the dissolution of a brand or com-pany relationship can leave customers with intense emotionsand cognitions that they express with overt behaviors towardformer relationship partners, just as they might in relationshipswith ex-friends or ex-lovers. For example, Marion, a long-termloyal customer of Yves Rocher (YR), expressed disappointmenton a Web forum, noting that she no longer buys YR products, aswell as some positive feelings, even after choosing to terminatethe relationship: “I still think a lot of my YR creams. For someI still haven’t found alternatives. They are the ones I missparticularly” (cited in Hemetsberger, Kittinger-Rosanelli, andFriedmann 2009, p. 435). On consumeraffairs.com, a consumer

0022-4359/$ – see front matter © 2010 New York University. Published by Elsevier Inc. All rights reserved.doi:10.1016/j.jretai.2010.09.004

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 401

Fig. 1. Post-termination responses as part of the relationship lifecycle.

named Michael expressed similar feelings: “the dealer of mybeloved BMW blew me off and I decided to quit this dealerand buy a Ford at another dealer. But I am still a bit homesickwhenever I see my beloved BMW. . ..”

As our main contribution, we introduce and explore a post-termination stage of customer relationships. This underexploredstage provides a bridge between the relationship dissolutionstage and the potential relationship recovery stage. Under-standing consumer responses in this post-termination stage canoffer companies substantial insights into the potential waysthey might revive terminated relationships. In addition, wecontribute to prior literature by establishing post-terminationcustomer heterogeneity and providing empirical evidence of dif-ferent customer clusters in the post-termination stage, which wecan identify according to differences in their post-terminationresponses. Finally, we explore how post-termination clusters dif-fer with regard to the reasons for their relationship termination.

In the remainder of this manuscript, we first outline the post-termination stage of consumer–brand relationships. To deriveinductive insights, we then employ a grounded theory approachand content analysis. We extract post-termination dimensionsusing optimal scaling and categorical principal component anal-ysis. In a subsequent cluster analysis, we use these dimensions toidentify four post-termination clusters and investigate the linksbetween reasons for termination and clusters with exact logisticregressions.

Conceptualizing post-termination responses

To introduce the concept of a post-termination stage ofconsumer relationships that bridges dissolution with potentialrevival, we investigate the consumer responses that define thisstage, namely the emotions, cognitions, and behaviors that con-sumers exhibit.

Current relationship research generally considers dissolutionas the final stage of a consumer relationship (Dwyer et al. 1987).Researchers identify determinants of relationship dissolution,such as changed needs, service failures, or variety seeking (e.g.,Hess, Ganesan, and Klein 2007; Keaveney and Parthasarathy2001; Seetharaman and Che 2009), but usually do not addresspost-termination responses. The only exceptions are Grégoireand Fisher (2008) who recognize that switched customers mightretaliate, and von Wangenheim (2005) who investigates post-switching negative word of mouth. Neither of these studies aimsat an in-depth analysis of consumers’ responses in the post-termination stage though.

As we illustrate in Fig. 1, the notion of post-terminationresponses extends the traditional relationship lifecycle andbridges the gap between relationship dissolution and revival ordetachment. We explore post-termination dimensions, use thesedimensions to identify post-termination consumer segments, anddistinguish these segments based on the termination reasons

Empirical context and procedure

This research focuses on premium car brands for two reasons:consumers typically maintain strong relationships with premiumcars (Kressmann et al. 2006; Parment 2008), and relationshipdissolution is usually a discrete event (i.e., the car buyer substi-tutes a competitor brand).3 To learn how consumers feel, think,and act after they separate from their former relationship part-ner, and considering the lack of research on post-terminationresponses, we adopted a qualitative grounded theory approach.

During in-depth interviews with consumers who had been ina strong relationship with their premium car brand but ended

3 Similarly intense relationships may appear in a wide range of other productsand industries (see Fournier 1998 for fast-moving consumer goods examples;see Gwinner, Gremler, and Bitner 1998 for service examples).

Author's personal copy

402 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413

that relationship and switched to a different brand, interviewersattempted to uncover subconscious motives and underlying psy-chological processes during the critical relationship terminationsituations while still remaining distant. The interviews were con-ducted by telephone. For the data analysis, we applied a methodtriangulation approach, combining qualitative and quantitativetechniques.

Sample

We conducted the empirical study in cooperation with amajor premium car producer (Brand A) that provided us accessto a sample frame of 1650 Dutch customers.4 We decided toinclude only customers in our sample who drove their cars fornon-business purposes and had terminated their relationshipwithin the previous four years. A longer time frame mightlimit customers’ memory of the relationship and reasons forterminating it.

Considering age, years of car ownership, and gender quotas,we called 1139 customers from the sample frame (three attemptswithin two weeks) and reached 673 of them, 403 of whom agreedto respond to a screening questionnaire (response rate ∼60%).Of the 403 consumers who responded, only 43 (10.7%) actu-ally met the brand relationship conditions; we conducted fullinterviews with all 43 consumers.

With regard to the sample demographics, 38 respondentswere men (88%), which is similar to the brand’s customer base.Respondents’ ages ranged between 31 and 78 years, with an aver-age of 58. On average, the telephone interviews lasted 65 min;the shortest interview was 40 min and the longest lasted 120 min.All respondents received a D 25 gift voucher in appreciation fortheir cooperation. The interviews were tape-recorded with therespondents’ approval and completely transcribed. We list keydemographic and usage-related information for each respondentin Appendix A.

Interview structure

We developed a semi-structured interview guide whichcontained open-ended questions addressing post-terminationresponses and reasons for ending the brand relationship. Wefirst asked respondents about their post-termination responses;to avoid the activation of unpleasant memories about the termi-nation, which might have influenced respondents’ subsequentanswers, the interviewers asked about their reasons for terminat-ing only after the respondents discussed their post-terminationstage responses. Finally, the questionnaire included closed-ended questions related to the respondents’ sociodemographicand household characteristics and the model of their most recentcar.

Within this structured interview process, interviewers alsocould adjust the questionnaire to match the specific responsesituation. For example, in addition to the interview guide, the

4 We substitute the actual brand name with “Brand A” to maintain the com-pany’s anonymity.

interviewers received a set of questions that would enable themto dig deeper into specific topics (e.g., “What exactly do youmean by . . ..?” “Could you please give an example?”).

Interviewers

The interviews were conducted by six psychology and/orinternational business graduate students. All interviewersattended two different training sessions that lasted three hourseach, which familiarized the interviewers with the idea of thepost-termination stage. A focused lecture summarized existingliterature in this research domain (session 1); in the role-playingsession, the interviewers practiced working with the interviewguide (session 2).

Data analysis

Grounded theory deconstructs and rearranges textual inputto identify categories and concepts and thus forms substantivetheory (Strauss and Corbin 1998). When analyzing the inter-views, we focused on the questions, “Which constructs play arole?” and “What processes take place?” to follow the standarddata analysis approach to grounded theory (Glaser and Strauss1967). In the context of our study, these questions refer to theconsumer’s cognitions, emotions, and behaviors after the ter-mination of a relationship with the focal brand, as well as theconsumer’s motivation to terminate the relationship.

We assigned codes to segments of text that represent identi-fying anchors for key information in the data (e.g., we coded therespondent statement, “Today everyone drives a Brand A. It is nolonger special,” as “negative communication about the brand”).We then generated concepts (e.g., “downgrading the brand”) bygrouping codes of similar content. By creating a typology inwhich we linked broad groups of similar concepts, we definedspecific categories (e.g., general typology of coping strategies).Finally, we derived explanations for the research subject (e.g.,emotional coping strategy).

Two psychologists with experience analyzing and codingqualitative studies conducted the coding and subsequent qual-itative data analysis. The data analysis was supported by thecomputer-assisted, qualitative data analysis software MAXqda(Weitzmann 2000) for cross-case representations, for which itprovided a systematic overview of all text that referred to specificthemes or constructs.

Quantitative methods complemented the qualitative results.Method triangulation offers a powerful approach rarely usedin retailing and marketing research (Olson 2004). We appliedoptimal scaling to quantify the constructs (Perreault and Young1980), which enabled us to conduct a categorical principal com-ponent analysis (CATPCA). The CATPCA results provided abasis for identifying post-termination clusters, and clusters weresubsequently used for studying the links between terminationreasons and segment membership with exact logistic regression.In the following sections, we present the termination reasons andthen the main post-termination responses we obtained from ourqualitative analysis of the interviews with previous car own-ers. We focus on the meaning of the identified post-termination

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 403

responses constructs and explicate them with illustrative cus-tomer quotes.

Reasons for relationship termination

Among respondents in our sample, relationships with the carcompany ended for either customer-related or brand/company-related reasons. Customer-related reasons comprised changingneeds and the customer’s financial reasons, generally unrelatedto brand performance. As Joseph (67 years) mourned: “After myheart attack I could no longer drive a car with gearshift,” andBarbara (75 years) noted: “Driving a Brand A car is fantastic.Reliable, but the entrance and the seats are hopeless. We talkeda lot about it, but it is simply the case. Otherwise we would havenever switched.”

If financial reasons led to the relationship termination, con-sumers indicated mainly that they could no longer afford thehigh-priced brand. According to Michael (42 years), “It wasnot an easy decision, but there was a price difference of 10,000Euro between the other brand and the latest Brand A model.” Theeffect of changing customer needs even might coincide financialreasons, as exemplified by Helen (51 years): “We now spend lotsof time in our Spanish house and hardly drive a car in the Nether-lands. I thought that having a smaller car in the Netherlandswould suffice.”

The brand/company-related reasons for termination referredto service delivery problems (Keaveney 1995) and core prod-uct failures. Customer relationships typically follow implicitcodes of conduct, according to which customers expect to betreated positively and with respect (Miller 2001). Yet severalrespondents indicated that they felt they had been treated ina disrespectful or anonymous way, despite their long personalloyalty history with the brand. Carl (72 years), a customerfor more than 40 years, described a recent experience: “Lasttime I was at the dealer, no one approached me. No one askedwhether s/he could help me. Then I decided that this was mylast time. I’ll never go there again.” Respondents who cited coreproduct failures as reasons indicated that they had to expendadditional investments of time and money, sometimes withouteven receiving an apology or understanding from the dealer: “Iat least expected something like ‘I am so sorry that you haveso many problems with your car, this is very inconvenient”’(Harold, 54 years).

Post-termination responses

Consistent with our expectations, we receive ample evidenceof intense and strong processing of the brand relationship inpost-termination stages. The interviews demonstrate that a sub-stantial share of consumers continue to display varied, strongcognitions, emotions, and behaviors in relation to the brand, evenafter they voluntarily dissolve their relationship. As we describenext, these consumer responses can be both negative andpositive.

Identifying post-relationship dimensions using CATPCA

Our discussion of the concepts that emerged from our stan-dard coding procedure follows the structure of a post hoccategorical principal component analysis (CATPCA), whichwe conducted by transforming the identified constructs intocategorical variables and assigning numerical values to the qual-itative data categories.5 As an exploratory nonlinear variation ofstandard principal component analysis (PCA), CATPCA usesoptimal scaling and can effectively deal with variables of mixedmeasurement levels that are not necessarily linearly related (e.g.,Gifi 1990; Perreault and Young 1980). It assigns optimal scalevalues to a variable’s categories to generate numeric-valuedtransformed variables (Linting et al. 2007), such that it candiscover category values that are optimal with a maximal over-all variance accounted for in the transformed variables, giventhe number of components. Moreover, CATPCA can deal withcategorical data and small samples, so it is well suited to our data.

We report the CATPCA results in Table 1. We extractedtwo post-termination dimensions with eigenvalues larger than 1,both of which have sufficient reliability for exploratory research(total Cronbach’s α = .95). This two-dimensional model of post-termination responses accounts for approximately 78% of thetotal variance. In Table 2, we list the various constructs weidentified in the qualitative interviews, along with their cate-gories according to the optimal scaling and their correspondingfrequencies.

Relationship-related responses

The first CATPCA dimension integrates four constructs withhigh positive loadings: attachment, identity, communication,and contact. These constructs constitute consumer responsesthat express how consumers think about, what they feel, orhow they behave toward the former brand, so we refer to thisdimension as “relationship-related responses.”

Attachment. Attachment theory from social psychology sug-gests that relationship partners develop strong bonds that maypersist even after separation (Bowlby 1973). Separation evokesdiverse, contrasting emotions including love, hate, bitterness,guilt, anger, envy, and concern. People who terminate a (social)relationship tend to miss each other after the termination, evenif the split was self-initiated (e.g., Clarke-Stewart and Brentano2006). Similarly, our interviews revealed both positive and nega-tive attachments to the former car brand, which differed in termsof their intensity.

Among the examples of positive attachment, positive mem-ories about the brand prevail. Peter (54 years) illustrates a stateof positive attachment: “I only had positive experiences withBrand A car. Both of us loved driving it.” Other respondentsexpressed an even stronger positive attachment, such that theyconsider the termination of the relationship a loss and express

5 This assignment resulted in dichotomous scaling (e.g., contact versus nocontact) for some constructs, whereas for others, the categorical labels weremore refined (e.g., positive, negative, ambivalent communication).

Author's personal copy

404 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413

Table 1Categorical principal component analysis results: two-dimensional model.

Construct Component/dimension

Relationship-related Termination-related

Attachment to former brand .968 .041Identity related to former brand .738 −.048Communication about former brand .867 −.210Contact with former brand .795 .031

Account making/attribution .119 .960Cognitive coping .374 .847Emotional coping −.434 .761

Dimension Cronbach’s α Variance accounted for

Total (eigenvalue) Percent of variance

1 .811 3.283 46.9032 .634 2.189 31.274

Total .953 5.472 78.178

Notes: Total Cronbach’s � is based on total eigenvalues. Values marked in bold indicate that a construct was assigned to the respective dimension.

Table 2Categorical constructs in the post-termination stage.

Construct/categories Frequency (N = 43)

Relationship-relatedAttachment to former brandVery positive 6Positive 20Neutral/dissociated attachment 7Negative 4Very negative 6Identity related to former brandInitial brand identity 17New brand identity 10Brand independent identity 16Communication about former brandPositive 19Ambivalent 11Negative 13Contact with former brandContact 24No contact 19Termination-relatedAccount making/attributionPersonal decision 24Dealer 10Brand 9Cognitive copingRational thinking 22Splitting 7No cognitive coping 14Emotional copingEmotional coping 16No emotional coping 27

feelings of homesickness, desire, and yearning for their favoredbrand. In the context of social relationships, Weiss (1975) labelsthis phenomenon “divorce pain,” described as a “response to theintolerable inaccessibility of the attachment figure” (p. 131). Forexample, Carl feels “homesick about my Brand A in general; italways used to be my car,” and Howard (55 years) reports, “My

heart is still with Brand A. My heart cries whenever I see a BrandA car passing by.”

In the socio-psychological context, Masheter (1991) findsthat divorced couples exhibit positively skewed post-divorceattachment scores; some former spouses even establish friend-ships after their divorce (Masheter 1997). Busboom et al.(2002) explain such friendship after divorce according tosocial exchange theory and argue that higher perceived bene-fits increase the likelihood of these friendships. Such positiveresponses after relationship termination may be common tosocio-psychological literature, but they represent a totally newphenomenon for retailing and marketing. The limited extant lit-erature on post-termination attitudes has reported only negativeresponses (Grégoire and Fisher 2008; von Wangenheim 2005).

Consistent with this extant literature, we also found nega-tive post-relationship attachments, again with various levels ofintensity. For example, James (41 years) believed, “They couldhave ensured that I would get a different feeling when hearingthe brand name Brand A.” But Harold had a stronger negativereaction: “Brand A used to be a status Brand, but I really don’tperceive this anymore. I think it absolutely does no longer haveany status.” If customers lack any interest in their former brandrelationship partner, they exemplify a neutral attachment (seeHowell 2008). For example, Kenneth (57 years) said plainly: “Ido not have a feeling toward the brand. It no longer touches me.”

Social identity. According to social identity theory (Tajfeland Turner 1986), recently divorced people can no longer referto an identity based on their spouse, so they strive to buildan independent identity. Similarly, respondents in our sampleexpressed concern about their changing identity after they hadterminated their relationship with the car brand. Specifically,we discerned three kinds of social identity effects in thepost-termination stage.

Some respondents continued to derive their identity, evenafter termination, from their former partner brand, a phe-nomenon we refer to as former brand identity. Fred (70 years)continued to assert that “My Brand A car was a car that matched

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 405

with me.” Other respondents instead distinguished themselvesto develop an identity related to their new brand (new brandidentity). As James mentioned: “My new brand is a bit smaller,a bit sportier, it fits better to me. I feel good in that car. Theentire [former] brand was not meant for me.” Finally, a thirdgroup of respondents do not rely on car-related relationshipsfor their identity (brand-independent identity), such as Kenneth,who argued: “[The car] is just an object, it does not touch me.”

Communication behavior. Our interviews support vonWangenheim’s (2005) finding of negative post-switching wordof mouth. In this spirit, James stated: “My relative consideredbuying a car of my former brand, but I discouraged him.” How-ever, our interviews also revealed some positive and ambivalentpost-termination communication. As an example of the former,David (66 years) told others that Brand A was “a reliable, highquality car, with a touch of luxury. Brand A really realizes salespromises about reliability and high quality.”

When they engaged in ambivalent communication, respon-dents stressed both positive and negative aspects of the brandin the same conversation, which indicated their conflicting feel-ings and thoughts about their former relationship. For example,Ralph (58 years), when asked what he tells others about BrandA, explained: “The new model X that I saw yesterday was reallybeautiful . . . and I still feel a little in love. Then I think, that’s anice car.. . . I think they should not have built model Y. . .. I wasreally fed up with it. . .. Way too much noise and then still fromBrand A, then I think, this should not be. That’s why I say, theyshould not have built it.”

Contact behavior. Just as divorced spouses exhibit differentcontact intensity than do spouses who remain married, cus-tomers who have terminated their brand relationship expresseddifferent levels of need to stay in touch with their brand.Consistent with our findings regarding attachment stylesand communication behaviors, respondents described differentcontact behaviors, including searching the Internet, trade publi-cations, and newspapers for information about the former brandpartner, such as “I read almost daily about Brand A, as I didtoday,” (Michael) and visiting car fairs: “Whenever I visit theAutoRAI [Dutch car fair] I will also visit Brand A as I simplyconsider it interesting” (Alan, 71 years). Other respondents ter-minated all contact with the brand. When asked whether he stillis in contact with Brand A, Henry (55 years) declared: “No. Therelationship has been terminated.”

Termination-related responses

The second CATPCA dimension combines the account mak-ing, cognitive coping, and emotional coping constructs, all ofwhich load highly and positively on the termination-relatedcomponent. Because the constructs all relate to perceptionsof relationship termination, we refer to this dimension as“termination-related responses.”

Account making. Account making includes explaining,describing, and emotionally responding to stress-inducing expe-riences, such as relationship termination; it constitutes anextension of attribution (e.g., Weiner 1985). Attribution mainlyfocuses on cognitive processes, but account making goes beyond

this cognitive perspective to include emotional and socialaspects, as well as sharing the story (Orbuch 1997). Accountscan be more effective than assigning blame as a means to pro-cess major life events. Weiss (1975) finds that account makinghelps people achieve closure in terminated social relationshipsand feel motivated to progress.

Respondents in our sample held either themselves or the com-pany responsible for the end of the relationship. Michael blamedhimself: “It was simply a financial decision I made. If I wouldhave had 10,000 Euro, I would have bought the new Brand A.”Ronald (64 years) instead developed a complete narrative andpresented the brand as a deceptive partner: “I bought my next[Brand A] as a new car and I have not been happy with it at all.Whenever I buy a brand new car, then I do not expect so manyproblems from a brand with a strong reputation.. . . When this isthe way things go at Brand A, this is no longer for me.”

Coping strategies. Coping refers to a person’s effort “to man-age specific external and/or internal demands that are appraisedas taxing or exceeding the resources of the person” (Folkmanand Lazarus 1988, p. 310). Most coping studies in social psy-chology refer to the stress people experience during divorce;in a brand relationship context, it refers to ways people handlethe stress they experience after the relationship with their brandhas ended. The interviews also suggested a distinction betweencognitive and emotional coping.

When consumers engage in cognitive coping, they emphasizecognitive domains (Duhachek 2005), such as rational thinkingand splitting. David, who would have loved to buy another BrandA, could not afford one after retiring and so used rational think-ing: “Whenever you know your own situation, you cannot beallowed to say ‘Brand A is the best car and therefore I mustbuy it.’ This is not the way it works in this world.” Walter (78years) instead used splitting, such that he separated the brand,the dealer, and the core product: “The dealer said that he con-tacted the brand and that the brand cannot do anything about it. Ihoped that the dealer would have said ‘this is our mistake; pleasecomply with the customer’s requests.”’ Furthermore, when theinterviewer added, “so this means that you are disappointed inthe way the brand treated you?” Walter firmly replied, “No! Iam not disappointed by the brand. I always had the feeling thiswas a mistake by the dealer.”

In contrast, emotional coping emphasizes personal feelingsabout experiences with the relationship partner (Folkman andLazarus 1988). Consumers who use an emotional coping strat-egy expressed contempt and adverse feelings toward the brand,such as when Frank (60 years) called his former dealer “arrogantmonkeys” and Harold demoted his former brand partner, suchthat “Brand A does not have any status at all anymore.”

Post-termination segmentation

Our discussion indicates strong differences across respon-dents in terms of their post-termination responses. By depictingthe biplots of respondents and CATPCA loadings to express therelationship between the identified constructs and the respon-dents, Fig. 2 points at the existence of distinct post-terminationcustomer clusters.

Author's personal copy

406 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413

Fig. 2. Biplots of respondents and CATPCA loadings.

To determine these post-termination clusters, we conducted apost hoc cluster analysis, using the standard squared Euclideandistance in combination with the Ward criterion. The two CAT-PCA scores were used as cluster variables. To determine theappropriate number of clusters, we used the pseudo-F indexdeveloped by Calinski and Harabasz (1974) which uses a vari-ance ratio criterion that puts the between-groups sum of squaresand the within-groups sum of squares into relation (large val-ues indicate a good solution), and the pseudo T-squared valuesresulting from the procedure suggested by Duda and Hart (1973),a ratio criterion that divides the within sum of squared errorsbefore a cluster merger by the squared errors after the merger(small values indicate a good solution). Both approaches werefound to determine the correct cluster number best in an exten-sive Monte Carlo study by Milligan and Cooper (1985). As canbe seen in Table 3, both methods suggested a four-cluster solu-tion; a result that was also consistent with the visual “elbowcriterion” (in which the number of clusters is plotted against thecumulative within-clusters sum of squares; Hair et al. 2006, p.594f.) and found to be both interpretable and manageable.

To develop a thorough understanding of the clusters, wereturned to the interview transcripts and reread them to com-bine the qualitative and quantitative insights. Therefore, whendescribing the clusters, we use the post-termination responses

(active cluster variables) and the termination reasons as wellas information about each cluster’s intention to repurchase, forboth the cluster as a whole and its centroid member (see Table 4for detailed profiles of the cluster centroids).

Cluster 1: Positive Attached (n = 18). Customers in this clus-ter terminated their relationship for personal reasons. Theycherished their positive memories and retained positive attach-ments to and identification with the former brand. They wouldtalk positively about the brand relationship and maintained con-tact. Approximately half of them intended to repurchase thebrand. Customer David, the cluster centroid, claimed he hadto end the brand relationship because he retired, but he still

Table 3Fit indices for different cluster solutions.

Number ofclusters

Pseudo-F index(Calinski and Harabasz 1974)a

Pseudo T-squared values(Duda and Hart 1973)b

2 30.82 98.063 95.27 27.664 114.38 8.155 110.53 16.01

Numbers in italics indicate best cluster solution.a Larger values indicate more distinct clustering.b Smaller values indicate more distinct clustering.

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 407Ta

ble

4Pr

ofile

ofcl

uste

rce

ntro

ids.

Clu

ster

1,Po

sitiv

eA

ttach

edD

avid

(66

year

s)C

lust

er2,

Hur

tYea

rner

sFr

ank

(66

year

s)C

lust

er3.

Furi

ous

Lea

vers

Har

old

(54

year

s)C

lust

er4,

Det

ache

dL

eave

rsB

rian

(48

year

s)

Hap

pym

arri

age

Dav

idw

asa

loya

lcus

tom

erfo

rab

out1

0ye

ars

driv

ing

two

cars

ofB

rand

Ain

row

.Thr

ough

outt

his

time

heha

da

very

posi

tive

bran

dre

latio

nshi

pw

ithB

rand

A.

Tho

ugh

heex

peri

ence

dco

refa

ilure

s,he

perc

eive

dth

emas

rela

tions

hip

ampl

ifier

sas

they

wer

eso

lved

succ

essf

ully

and

heex

peri

ence

dgo

odw

illby

the

bran

d(n

opa

ymen

tfor

afa

ilure

even

afte

rgu

aran

tee

endi

ng)

and

bene

vole

ntde

aler

trea

tmen

te.g

.by

getti

nga

rent

alca

rfo

rfr

eedu

ring

the

repa

ir.(“

Inth

ebe

ginn

ing

Iw

asdi

sapp

oint

ed,b

utth

enI

saw

how

itw

asso

lved

and

this

chan

ged

my

feel

ing.

[...

]it

was

solv

edpr

oper

lyan

dw

ithou

tany

char

ge.”

).

Fran

kha

sbe

enan

enth

usia

stic

and

loya

lcus

tom

erof

Bra

ndA

(“I

used

tobe

alo

yalb

uyer

and

thri

lled

driv

er.

Iha

vene

ver

hidd

enm

ypo

sitiv

eex

peri

ence

s”).

The

last

year

sFr

ank

has

been

driv

ing

the

mos

texp

ensi

veB

rand

Aca

ran

dhe

isve

rypr

oud

abou

tit(

“you

shou

ldkn

owth

atI

amve

rypr

oud

ofm

yx-

mod

el”)

.He

talk

sfu

llof

enth

usia

smab

outh

isha

ppin

ess

with

the

bran

d(“

Iw

asth

rille

dab

outm

yx-

mod

el,a

ndon

ceI

had

axy

whi

chm

ade

me

soha

ppy”

).Fr

ank

isab

leto

expr

ess

his

stat

usan

dpo

rtra

yhi

mse

lfby

driv

ing

Bra

ndA

.For

him

Bra

ndA

isa

repr

esen

tativ

eca

rw

hich

regu

larl

yof

fers

him

reco

gniti

on.(

“With

the

xm

odel

,one

can

impr

ess.

Itis

ave

ryre

pres

enta

tive

car.

And

ever

ythi

ng,e

very

thin

gw

asgo

od.W

hen

Iar

rive

som

ewhe

re,I

freq

uent

lyre

ceiv

epo

sitiv

ere

actio

nsab

outm

yca

r.”)

Har

old

used

todr

ive

Bra

ndY

for

man

yye

ars

befo

rebu

ying

Bra

ndA

for

the

first

time.

“Itw

asal

way

sa

drea

mto

driv

eB

rand

A.”

Inth

ebe

ginn

ing

ofth

ere

latio

nshi

phi

sbr

and

perc

eptio

nw

as“a

nim

age

ofbe

ing

very

solid

,rel

iabl

ean

dw

itha

high

resa

leva

lue.

”B

ased

onhi

spe

rcep

tion

ofa

fair

rela

tions

hip,

Har

old

was

will

ing

tobe

arth

eim

bala

nce

duri

ngth

ere

latio

nshi

pw

ithou

tany

com

plai

nts

(“I

did

notc

ause

any

trou

ble

desp

item

ydi

sapp

oint

men

t”)

assu

min

gth

atth

ede

aler

wou

ldre

stor

eth

eba

lanc

ein

case

ofa

repu

rcha

se.

Bri

andr

ove

two

diff

eren

tmod

els

ofB

rand

Afo

rab

outf

our

year

s.In

Bri

an’s

perc

eptio

n,B

rand

Aen

joys

ago

odre

puta

tion

ofm

anuf

actu

ring

dura

ble

and

relia

ble

cars

.

Rea

sons

tote

rmin

ate

the

rela

tions

hip

The

only

reas

onfo

rD

avid

tote

rmin

ate

the

bran

dre

latio

nshi

pw

asdu

eto

pers

onal

reas

ons.

He

wou

ldha

velo

ved

toco

ntin

ueth

ere

latio

nshi

pw

ithB

rand

A,

butd

ueto

his

retir

emen

tand

his

pref

eren

cefo

ran

auto

mat

icge

arB

rand

Aw

asno

taff

orda

ble

anym

ore

([..

.]“a

nau

tom

atic

gear

from

Bra

ndA

isab

ove

my

retir

emen

tpay

,let

me

puti

tthi

sw

ay.T

here

fore

,we

chan

ged

toa

chea

per

car

and

ende

dup

atbr

and

X.”

).

Fran

k’s

first

disa

ppoi

ntm

enti

nth

ese

rvic

ein

tera

ctio

nst

arte

dw

ithth

ela

stpu

rcha

seof

Bra

ndA

.Whe

nFr

ank

pick

edup

his

new

car,

the

deal

erm

ade

fun

abou

tthe

car

(“th

ede

aler

told

me

Ish

ould

notd

rive

too

fast

,be

caus

eth

em

otor

coul

dfa

llou

tand

mor

eof

thes

e‘j

okes

’,w

hich

frus

trat

edm

ea

lot”

).Y

ears

late

rw

hen

Fran

kw

aspl

anni

ngan

othe

rpu

rcha

seof

ane

wca

r,he

wou

ldlik

eto

cont

inue

the

rela

tions

hip

and

buy

agai

non

eof

the

mos

texp

ensi

veca

rs.T

hede

aler

how

ever

does

nots

eem

tobe

inte

rest

edin

are

latio

nshi

pw

ithFr

ank

and

expr

esse

sth

athe

isju

ston

eof

man

ycu

stom

ers.

Fran

kex

peri

ence

sth

atth

ere

latio

nshi

pis

nota

bala

nced

part

ners

hip,

butr

athe

ran

asym

met

ric

pow

erre

latio

nshi

pin

whi

chhe

isde

pend

entu

pon

the

deal

er.I

nco

ntra

stto

the

past

the

deal

eris

notw

illin

gto

disc

uss

the

resa

lepr

ice

ofth

eus

edca

r.D

espi

tepo

sitiv

eat

tach

men

tand

satis

fact

ion

abou

tthe

car,

Fran

kde

cide

dto

term

inat

eth

ere

latio

nshi

pin

volu

ntar

ily,a

sth

ede

aler

seem

edto

abus

ehi

spo

wer

and

can

nolo

nger

bese

enas

afa

irpa

rtne

r.

Dur

ing

thei

rre

latio

nshi

pw

ithB

rand

A,c

usto

mer

sex

peri

ence

ddi

ffer

entk

inds

ofst

rong

,uns

olve

dco

refa

ilure

s,m

any

times

acco

mpa

nied

byad

ditio

nals

ervi

ceen

coun

ters

.D

urin

gth

etim

eH

arol

dpo

sses

sed

the

car,

heex

peri

ence

dm

any

prod

uctf

ailu

res

(“I

had

som

uch

trou

ble

with

this

thin

g.Y

ouev

enw

ould

noth

ave

acce

pted

this

from

ach

eape

rbr

and”

).T

here

latio

nshi

pbe

com

esw

orse

,whe

nH

arol

dex

peri

ence

sth

atth

ede

aler

does

noti

nten

dto

offe

ran

yre

cove

ryac

tion.

“You

real

lyfe

elbe

tray

ed.I

fyo

ukn

owth

ata

cust

omer

has

had

som

uch

bad

luck

,the

nyo

ual

sokn

owth

atth

esm

alle

stpr

oble

mdu

ring

the

repu

rcha

sesi

tuat

ion

will

caus

ea

brea

kup

[...

]an

dw

hatm

akes

me

mos

tun

happ

yw

asth

ere

purc

hase

.Thi

sbo

ther

edm

eso

muc

h.T

here

,Ire

ally

had

the

feel

ing

now

Ige

trip

ped

off.

The

trad

e-in

pric

ew

assc

anda

lous

low

and

they

knew

it.E

ven

aye

arla

ter,

the

deal

erof

bran

dY

offe

red

atr

ade-

inpr

ice

of25

%m

ore.

Aye

arla

ter

and

ata

diff

eren

tbra

nd.”

Dur

ing

thei

rre

latio

nshi

pw

ithB

rand

A,c

usto

mer

sex

peri

ence

mor

eor

less

seve

rese

rvic

een

coun

ters

and/

orco

refa

ilure

s.A

dditi

onal

ly,p

erso

nal

rest

rict

ions

affe

ctfu

ture

affo

rdab

ility

.C

onsi

sten

twith

this

perc

eptio

n,B

rian

boug

hta

dies

elve

hicl

e,w

hich

caus

edhi

mtr

oubl

e,w

hen

usin

gth

eca

rfo

rci

tyri

des.

(“I

just

driv

ein

the

city

,so

itw

asa

wro

ngpu

rcha

se.I

shou

ldha

vebo

ught

ape

trol

car,

buto

neso

met

imes

sim

ply

does

thin

gs.

AtB

rand

Aon

edr

ives

adi

esel

car

and

nota

petr

ol.”

).H

een

dsth

ere

latio

nshi

pbe

caus

eof

the

core

failu

res

(“T

heca

rw

asm

ore

inth

ega

rage

than

anyt

hing

else

.And

that

was

mor

eor

less

the

reas

onto

getr

idof

it”)

and

beca

use

ofhi

sri

sing

awar

enes

sof

his

finan

cial

limita

tions

(Ith

ough

t“it

isa

too

expe

nsiv

eca

rfo

rm

e.”)

.

Post

-rel

atio

nshi

ppr

oces

ses

Rel

atio

nshi

p-re

late

dre

spon

ses

Atta

chm

ent

Dav

idst

ates

“atta

chm

enti

sa

far

too

stro

ngw

ord

for

aca

rbr

and,

”,bu

the

has

“pos

itive

emot

ions

tow

ards

Bra

ndA

[...

]ba

sed

onpo

sitiv

eex

peri

ence

s”an

dsa

ys“I

have

noty

etsa

idgo

od-b

yeto

Bra

ndA

.”).

Dis

tingu

ishi

ngbe

twee

nth

ego

odbr

and

and

the

bad

deal

eral

low

sFr

ank

toco

ntin

uehi

sst

rong

atta

chm

ent

toth

eex

istin

gB

rand

A.F

orFr

ank,

Bra

ndA

isst

illth

esu

peri

orpr

oduc

t(“B

rand

Ais

solid

and

Bra

ndB

isa

very

nice

car.

But

ifyo

ucl

ose

the

door

from

Bra

ndA

and

Bra

ndB

,tha

tis

adi

ffer

ence

ofni

ghta

ndda

y.”)

.

Har

old

show

sst

rong

nega

tive

atta

chm

enti

mpl

icitl

yby

reje

ctin

gth

epr

evio

usim

port

ance

ofth

ebr

and

and

his

“neg

ativ

eem

otio

ns”

tow

ards

the

deal

er.

Iden

tity

Dav

idsh

ows

his

iden

tifica

tion

with

Bra

ndA

due

tohi

sbr

and

enha

ncin

gco

mm

ents

whi

lene

glec

ting

his

curr

entb

rand

rela

tions

hip

with

bran

dX

.

Fran

kco

nsid

ers

his

form

erre

latio

nshi

pw

ithB

rand

Asu

peri

orto

the

new

bran

dre

latio

nshi

pw

ithbr

and

B(“

Bra

ndA

isst

rong

er.T

his

was

the

case

and

will

alw

ays

beth

eca

se),

the

bran

dm

eani

ngan

dpo

sitiv

ebr

and

rela

tions

hip

cont

inue

sev

enaf

ter

rela

tions

hip

term

inat

ion.

Tho

ugh

hein

tend

edto

cont

inue

the

rela

tions

hip

with

Bra

ndA

,Har

old

stat

esth

athe

iden

tifies

with

his

prev

ious

lydr

iven

bran

dY

whi

chhe

now

driv

esag

ain

(“T

his

isa

car

whi

chfit

sm

e.[.

]T

his

stub

born

ness

ofbr

and

Yw

asal

way

sap

peal

ing

tom

e.”

Author's personal copy

408 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413Ta

ble

4(C

onti

nued

)

Clu

ster

1,Po

sitiv

eA

ttach

edD

avid

(66

year

s)C

lust

er2,

Hur

tYea

rner

sFr

ank

(66

year

s)C

lust

er3.

Furi

ous

Lea

vers

Har

old

(54

year

s)C

lust

er4,

Det

ache

dL

eave

rsB

rian

(48

year

s)

Com

mun

icat

ion

Dav

id’s

com

mun

icat

ion

abou

tthe

bran

dor

part

sof

the

bran

d(i

.e.p

rodu

ct,d

eale

r)is

posi

tive.

He

desc

ribe

shi

scu

rren

tbra

ndpe

rcep

tion

as“a

relia

ble,

qual

itativ

ehi

gh-e

ndca

rw

itha

touc

hof

luxu

ry[.

..]

and

this

isno

ton

lysa

les

rep

talk

abou

trel

iabi

lity

and

qual

ity,B

rand

Ado

estr

ansl

ate/

impl

emen

tthi

sin

thei

rca

rs).

”

Fran

k’s

invo

lunt

ary

endi

ngof

the

rela

tions

hip

isac

com

pani

edby

grie

fab

outt

helo

ssof

the

rela

tions

hip.

He

says

:“I

amsa

tisfie

dw

ithth

isne

wbr

and,

alth

ough

som

ethi

ngis

mis

sing

:Iti

sno

tBra

ndA

.”In

the

com

mun

icat

ion

abou

tthe

bran

d,Fr

ank

expr

esse

sbr

and-

stre

ngth

enin

gan

dbr

and

harm

ing

lexi

s.T

hepe

jora

tive

rem

arks

can

bese

enas

emot

iona

lcop

ing

with

the

rela

tions

hip

term

inat

ion.

The

exis

ting

posi

tive

atta

chm

entt

owar

dsth

ebr

and

isre

flect

edin

posi

tive

com

mun

icat

ion.

For

Fran

k,th

epo

sitiv

eco

mm

unic

atio

nal

sosh

ows

inhi

sre

com

men

datio

nbe

havi

or(“

my

son

inla

wus

edto

driv

ebr

and

C,n

owhe

driv

esB

rand

A.S

oI

still

have

this

muc

hin

fluen

ce(l

augh

s).”

Inlin

ew

ithH

arol

d’s

emot

iona

lcop

ing

his

com

mun

icat

ion

abou

tthe

bran

dis

nega

tive.

Bri

ando

esno

tdev

alue

Bra

ndA

byhu

mili

atin

gor

belit

tling

the

bran

dor

othe

rcu

stom

ers,

buth

eis

criti

cali

nhi

sco

mm

ents

abou

tthe

prod

uct(

“For

such

anex

pens

ive

car

itis

rath

ersm

all”

)an

dun

sure

,whe

ther

hew

ould

reco

mm

end

Bra

ndA

(Bri

an(i

ndo

ubt)

:“I

dono

tkno

w,t

here

are

som

any

good

cars

.”).

Con

tact

Dav

idis

still

inco

ntac

twith

the

bran

dan

dis

follo

win

gth

eir

curr

entd

evel

opm

ents

(“W

hen

Ise

ea

new

spap

erar

ticle

abou

tBra

ndA

,Iam

alw

ays

inte

rest

edin

it[.

..]

and

soI

amst

illin

cont

actw

ithB

rand

A.”

)

Des

pite

the

ambi

guou

sco

mm

unic

atio

n,Fr

ank’

sow

nin

form

atio

nse

arch

beha

vior

lead

sto

cont

actt

oth

efo

rmer

bran

d.

Whe

nas

ked

abou

this

inte

rest

inB

rand

A,H

arol

dre

plie

s:“n

o,ne

ver

ever

.”H

eis

noti

nter

este

din

bran

dde

velo

pmen

tsan

ddo

esno

twan

tto

stay

inco

ntac

twith

the

bran

d(I

nter

view

er:“

Are

you

still

inte

rest

edin

Bra

ndA

cars

?”B

rian

:“(s

ilenc

e)H

mm

mm

.No.

”In

terv

iew

er:“

No?

”B

rian

:“N

o.”

Inte

rvie

wer

:“D

oyo

use

arch

for

info

rmat

ion

abou

tBra

ndA

cars

ordo

you

read

abou

tthe

m?”

Bri

an(w

ithem

phas

is):

“No,

no.”

)Te

rmin

atio

n-re

late

dre

spon

ses

Attr

ibut

ion/

acco

untm

akin

gIn

his

acco

untm

akin

g,D

avid

pres

ents

him

self

asre

spon

sibl

efo

rth

ere

latio

nshi

pte

rmin

atio

n,w

here

asth

ebr

and

ispr

esen

ted

asa

supp

ortin

gre

latio

nshi

ppa

rtne

r,w

hotr

ied

tom

ake

itpo

ssib

leto

cont

inue

with

the

rela

tions

hip

(“H

e[t

hede

aler

]of

fere

dm

ea

good

deal

[...

]a

good

trad

e-in

pric

efo

rm

yca

r,m

ore

than

Ifin

ally

gotf

rom

Bra

ndX

.”).

Aft

erre

latio

nshi

pte

rmin

atio

n,in

his

acco

untm

akin

g,Fr

ank

blam

esth

ede

aler

for

the

diss

olut

ion.

Inth

epo

st-r

elat

ions

hip

term

inat

ion

stag

e,H

arol

dm

akes

acco

untb

ypr

esen

ting

him

self

asth

evi

ctim

and

the

bran

das

ade

cept

ive

and

non-

bene

vole

ntre

latio

nshi

ppa

rtne

r,w

hoon

lyw

ants

toca

pita

lize

onits

cust

omer

s.

Inhi

sac

coun

tmak

ing,

Bri

angi

ves

ate

chni

cal

expl

anat

ion,

why

adi

esel

engi

neha

sto

brea

k,if

you

only

driv

ein

the

city

.(“I

gote

very

time

glow

plug

san

dlit

tlete

chni

calf

ailu

res

beca

use

ofth

esh

ortr

ides

Im

ade.

And

that

had

todo

with

the

pres

sure

diff

eren

ces

that

occu

r.A

ndth

ose

occu

rw

hen

mak

ing

shor

trid

es.”

).B

yth

iski

ndof

expl

anat

ion

heon

lybl

ames

him

self

,and

pres

ents

bran

dan

dpr

oduc

tas

bein

gno

tres

pons

ible

.As

the

Det

ache

dL

eave

rsfe

elre

spon

sibl

efo

rth

eir

switc

hing

,the

ydo

notf

eelt

reat

edba

dly

nor

reje

cted

byth

ebr

and

(or

deal

er).

Cop

ing

Dav

idco

pes

with

his

rela

tions

hip

term

inat

ion

byra

tiona

lizin

gan

dpu

tting

ratio

nal-

logi

calm

otiv

esin

the

fron

t(“S

ee,t

hose

are

situ

atio

nsw

here

you

only

can

deci

deif

you

know

your

pers

onal

situ

atio

n.T

hen,

you

cann

otal

low

your

self

tosa

y:‘B

rand

Ais

the

best

car,

henc

eis

has

tobe

Bra

ndA

’.It

does

notw

ork

like

this

inou

rw

orld

.”).

Fran

kap

plie

sem

otio

nalc

opin

gby

deva

luin

gth

epr

oduc

t(“t

hene

wm

odel

ofB

rand

Ais

notw

orth

the

mon

ey”)

and

the

deva

luin

gth

ede

aler

netw

ork

(“th

ede

aler

sar

ere

puls

ive”

,“ar

roga

ntm

onke

ys”)

.Fra

nkcl

earl

ydi

stin

guis

hes

betw

een

the

bran

dan

dth

ede

aler

(“B

rand

Ais

notr

espo

nsib

le,I

assu

me

the

deal

erha

shi

sow

nre

spon

sibi

lity”

)w

hich

can

bese

enas

aty

peof

cogn

itive

copi

ng.

Har

old

cope

sw

ithth

ebr

and

reje

ctio

nan

dex

ploi

tatio

nby

deva

luin

gth

ebr

and

itsel

f(“

Bra

ndA

has

nost

atus

atal

lany

mor

e.”)

and

accu

sing

itto

bea

liar

(“B

rand

Apr

omis

esqu

aliti

es,w

hich

–as

Ire

aliz

ed–

are

not

true

.”).

Tho

ugh

hein

tend

edto

buy

anot

her

car

ofB

rand

A,i

nth

epo

st-r

elat

ions

hip-

stag

ehe

pres

ents

the

curr

entp

rodu

ctof

Bra

ndA

ina

very

nega

tive

man

ner

(“di

sapp

oint

ing

qual

ity”,

“cos

tly,”

“not

inno

vativ

e,”

“not

real

lyou

tsta

ndin

g,”,

“mod

eldi

dno

tcha

nge,

”“B

rand

Aus

edto

bele

adin

gin

safe

tyte

chno

logy

,now

they

are

pass

edby

any

med

ium

-siz

edca

rbr

and

asth

ose

offe

rbe

tter

safe

tyte

chno

logy

”),a

way

ofem

otio

nalc

opin

g.

Tho

ugh

Bri

andr

ove

Bra

ndA

two

times

inro

w,h

efin

dsa

ratio

nale

xpla

natio

n,w

hyit

isno

tap

prop

riat

eto

driv

eB

rand

Aan

ymor

e(“

Iha

dth

efe

elin

gth

atm

ycu

stom

ers

have

topa

yfo

rmy

car”

).W

hen

talk

ing

abou

tthe

bran

d,he

does

nots

how

any

emot

ions

nor

uses

emot

iona

lter

ms.

Tho

ugh

Bri

anm

ainl

ydr

ives

prem

ium

bran

ds-w

hich

isno

rmal

lyac

com

pani

edby

emot

ions

-he

redu

ces

the

car

to“

Itis

just

am

ean

for

tran

spor

tati

on.”

Alth

ough

heon

cepu

rcha

sed

Bra

ndA

,he

now

mak

escl

ear,

that

hedo

esno

tbel

ong

toits

cust

omer

grou

p(“

Loo

k,w

hen

Iam

adi

rect

orof

aco

mpa

nyof

20pe

ople

and

Iha

veto

driv

eth

roug

hE

urop

eal

lwee

k,th

enI

wou

ldde

finite

lydr

ive

Bra

ndA

.But

Iam

not.

And

Ido

noth

ave

toso

.”)

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 409

thought, felt, and talked positively about the brand, which hekept contact with by reading newspaper articles. He practicedmainly cognitive coping.

Cluster 2: Hurt Yearners (n = 8). This cluster exhibitedstrong responses related to both post-termination dimensions.Members terminated their brand relationship in response tocompany-related service delivery failures and engaged in emo-tional coping. Yet they still revealed positive or very positiveattachments and identified with their former brand. They spokepositively and negatively about their brand relationship after itended; almost all cluster members maintained contact with thebrand by gathering information. About half of them indicatedthey would favor Brand A for their next purchase. Cluster cen-troid Frank was hugely disappointed by but still identified with(and was attached to) the brand. He felt a sense of grief aboutthe end of their relationship, so he engaged in strong emotionalcoping to develop his feelings toward both the brand and thedealer.

Cluster 3: Furious Leavers (n = 12). Customers in this clus-ter left because of core product or service failures. They blamedthe brand for the end of their relationship and used emotionalcoping. Their relationship-related responses were very negative;consumers still suffered from frustration and negative emo-tions. Negative attachment and communication dominated, andthey developed identities related to a new brand. None of themwould consider Brand A for their next car purchase. Cluster cen-troid Harold experienced several product failures and a sense ofbetrayal by the brand dealer; he remained strongly negativelyattached and identified with a competitive brand. Harold con-sidered himself a victim and coped emotionally by devaluingthe brand.

Cluster 4: Detached Leavers (n = 5). Finally, the members ofthis cluster are the only ones who expressed neither relationship-nor termination-related responses. In a sense, these consumersreflect the traditional marketing thinking that considers rela-tionship termination the final stage of the relationship, withvery limited post-termination processing taking place. We find

it insightful that this cluster is the smallest group in our sam-ple. These customers all took responsibility for what happenedand used cognitive coping strategies. Respondents no longersensed an attachment to Brand A; instead, they derived theiridentity from a new brand relationship. Consistent with thislack of attachment, they were no longer interested in the brandand rejected any ideas about recovering the relationship. Cen-troid Brian (48 years) terminated his brand relationship afterseveral failures, but did not blame the brand, because accord-ing to his cognitive coping strategy the reason for the failureswere “technical.” Brian also maintained no contact with thebrand.

Linking termination reasons and post-termination clusters

To deepen our understanding of the link between the post-termination clusters and the reasons that led members toterminate their brand relationships, we apply exact logisticregression analysis (ELR) (Hirji, Mehta, and Patel 1987; Mehta1995). When traditional asymptotic methods for analyzing datasets become unreliable because of small sample sizes such as theone used herein, ELR offers an important analytical alternative.Because the reasons for relationship termination did not serve asactive cluster variables in the segmentation, ELR also providesa test of external validity (Dant and Gundlach 1998).

We compared each cluster against the remaining sample.As we report in Table 5, termination reasons have significantimpacts on cluster allocation in all cases. By testing whether thereason parameters equal zero (H0: β1 = β2 = β3 = 0, or the jointoption), we can reject H0 for all four clusters. The results alsosupport the external validity of the cluster solution.

Personal reasons (p < .001) mainly distinguish the positiveattached (Cluster 1) from the remaining sample. Apparently,the personal decision to terminate the relationship results inpositive post-termination responses toward the former brand.Service encounter failures (p < .001) instead distinguish the hurtyearners (Cluster 2); in their post-termination responses, these

Table 5Exact logistic regression results.

Dependent variable (cluster comparison) Independent variable (reasons to terminate) Coefficienta (p-value) Odds ratiob

Positive attached Joint (0.000)Personal reason 4.932 (0.000) 138.661Service encounter 0.080 (0.480) 1.083

Hurt yearners Joint (0.000)Service encounter 3.511 (0.000) 33.480Core failure 0.894 (0.290) 2.444

Furious leavers Joint (0.000)Service encounter 2.525 (0.011) 12.491Core failure 4.696 (0.000) 109.555

Detached leavers Joint (0.000)Personal reason −3.715 (0.001) 0.024Service encounter −4.015 (0.001) 0.018Core failure −3.724 (0.001) 0.024

a Median unbiased estimates (MUE).b An odds ratio higher than 1.0 indicates that that specific reason to terminate the relationship has greater odds in the reported cluster than in the rest of the sample;

an odds ratio lower than 1.0 means that a reason to terminate the relationship has smaller odds in the reported cluster than in the rest of the sample.

Author's personal copy

410 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413

Table 6Switch reasons and post-termination responses per cluster.

Cluster 1: positiveattached

Cluster 2: hurt yearners Cluster 3: furious leavers Cluster 4: detached leavers

Switch reason Personal Company (service failures) Company (coreproduct/service failures)

Combination of personaland company

Repurchase intentions Yes Yes No No

Relationship-related responsesAttachment to former brand Positive Positive Negative NeutralIdentity related to former brand Initial brand Initial brand New brand New brandCommunication about former brand Positive Ambivalent Negative AmbivalentContact with former brand Yes Yes No No

Termination-related responsesAccount making (blaming) Personal decision Dealer Brand Personal decisionCognitive coping Rationalization No No RationalizationEmotional coping No Yes Yes No

customers blame the dealer and cope emotionally with the per-ceived humiliation, but they still exhibit positive responses tothe brand. A combination of service encounter failures (p < .05)and core failures (p < .001) separated the furious leavers (Cluster3) from the remaining sample, which likely explains their nega-tive post-termination responses and tendency to blame the brandfor the relationship dissolution. Finally, the detached leavers(Cluster 4) differ from the other clusters with regard to per-sonal reasons (p < .001), service encounter failures (p < .001),and core failures (p < .001), with all three termination reasonshaving smaller odds to occur in isolation in this cluster thanin the rest of the sample. However, a separate logit regressionincluding the combined occurrence of personal and companyreasons as independent variable (something that was exclusiveto this cluster and affected four the five cluster members) showeda highly significant positive coefficient of 1.665 for this variableand an odds value of 92.25. We summarize the four clusters inTable 6.

Discussion and implications

Implications for theory and further research

We introduce and explore the post-termination stage ofconsumer relationships, an underresearched link between thestages of relationship dissolution and potential relationshiprecovery. Our investigation provides empirical evidence that asubstantial share of customers undertake intense and complexpost-termination processing with regard to both the relation-ship and its termination. As another substantive contribution,we empirically account for post-termination heterogeneity byidentifying four illustrative post-termination clusters.

Relationship-related responses encompass attachment, iden-tity, communication, and contact. The few existing studies onpost-switching behavior mainly report negative responses, butwe find that about 60% of the respondents in our sample exhibitpositive or very positive attachment to the brand even after rela-tionship termination; members of clusters 1 (Positive Attached)and 2 (Hurt Yearners) show a particularly high level of pos-itive attachment, while members of 3 (Furious Leavers) and

4 (Detached Leavers) are less positively attached.6 Attach-ment thus represents a crucial concept for research on recoverymanagement (e.g., Stauss and Friege 1999). Given the high-involvement context of this research, scholars are encouraged toconsider whether this finding holds in a low-involvement settingtoo.

The identity construct also emerged as a key relationship-related response; 40% of our respondents still identified withtheir former brand (identification was particularly high for “HurtYearners” and “Positive Attached” segments), and 44% pro-vided positive communication about it (with “Positive Attached”members engaging strongest in communicative activities). Thisfinding extends Fournier’s (1998) discussion of the role of iden-tity in brand relationships to the post-termination stage; that is,identification may last longer than the relationship. A promisingarea for study includes the assessment of causal relationshipsamong these post-termination responses.

The final relationship-related construct refers to customers’contact behavior. We find that 56% of respondents still had somekind of contact with their former brand (contact being high-est for “Hurt Yearners”). Additional research should investigatewhich relationship revival strategies would be most effective forcustomers with high contact across different contact points.

Furthermore, we identify termination-related responsesamong the 23% of respondents who attribute blame to thedealer (mostly members of “Hurt Yearners”) and 21% whoblame the brand (mostly “Furious Leavers”). Respondents whoblame the brand are far more negative in their post-terminationresponses than are respondents who blame the dealer; brandattachment appears to be stronger than disappointment in thedealer. Spillover effects between brand and dealer attributionsthus would be an intriguing area for further research. Respon-dents also cope either rationally or emotionally with the loss oftheir consumer–brand relationship.

6 Although being informative about the responses’ relevance for our respon-dents, please note that the sample percentages reported in this section should notbe interpreted as representative of the population in general, or of other productcategories, as a result of the small sample size.

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 411

When linking termination reasons and cluster membershipwith an exact logistic regression, we found that termination rea-sons provide a solid basis for predicting a customer’s clustermembership, though clearly more work is needed in this area.Additional studies should combine existing work on recoverymanagement with our customer clusters to investigate whichare most receptive to win-back efforts.

Retail implications

A fundamental implication of this research is that retail-ers and other companies should account for the existence ofa post-termination stage of customer relationships when design-ing CRM programs. To validate this claim, we conducted threepost hoc interviews with managers from the automotive indus-try, which all three saw strong business potential in the extensionof CRM activities toward customers who have reached theirpost-termination stage (all three managers’ companies were notacting on post-termination stage customers).7

Our research provides companies with knowledge on theinternal and overt processes that take place in the post-termination stage and offers a first look at inter-customerdifferences. Our respondents noted that the lack of knowl-edge was the main reason for not systematically exploring thebusiness potentials provided by the post-termination stage. Toexploit the knowledge generated by this research, it will becrucial for companies to systematically collect information ontermination reasons and post-termination behaviors. An impor-tant role in this regard can be assigned to salespeople who shouldpay attention to lost customers who still maintain contact withthe brand, such as visiting the brand at a car fair. Other mediawhich offer powerful information might include the customer’scontinuing newsletter membership or his or her contributions tothe brand’s (online) community.

Such information should then be used to assign a formercustomer to one of the four post-termination segments and todevelop appropriate measures for re-vitalizing terminated rela-tionships. Again, the salespeople might help, as manager Aargues: If a lost customer with positive feelings toward the brandvisits the car fair, sales representatives can propose an appro-priate offer to entice the customer to return to the brand. Formembers of clusters 3 and 4, similar proactive behavior (e.g.,offering a better trade-in price, manager B) can help avoid nega-tive word-of-mouth communication, another benefit of engagingin post-termination marketing activities.

Finally, we encourage companies to include the post-termination stage in their relationship lifecycle models.Combining insights on such lifetime models with post-termination knowledge, dealers might want to contact not onlyexisting, but also “lost” customers (at least those of clusters 1and 2) before their finance contract with the new brand is about

7 Respondent A was a marketing manager of a car producer (male, 36 years);Respondent B represented a car dealer (male, 57 years), and respondent C man-aged an automotive supplier (male, 43 years). All three managers lived in theNetherlands when the interviews were conducted.

to end (in the car industry, this is typically the case after 25–37months; manager C).

In summary, marketers can benefit from acknowledgingthe post-termination stage and managing corresponding cus-tomer responses effectively. Retailers should be aware ofpost-termination heterogeneity, which has key implications forcustomer recovery management.

Appendix A. Sample description

Object Nr Name Gender Age Monthssinceswitch

Householdsize

Last carmodel

101 John Male 42 9 4 Smallfamily

102 Harry Male 64 9 2 Smallfamily

103 Paul Male 48 8 2 Compact104 Barbara Female 75 18 2 Compact105 Louis Male 44 12 2 Luxury106 Howard Male 55 21 1 Compact107 Alan Male 71 40 2 Compact108 Linda Female 70 15 2 Compact109 David Male 66 14 2 Small

family110 Thomas Male 59 17 2 Compact111 Helen Female 51 9 3 Compact112 Thomas Male 39 21 4 Compact113 Charles Male 70 11 2 Compact114 Michael Male 42 9 3 Compact115 Gary Male 49 18 1 Luxury116 Tim Male 35 15 5 Compact117 Joseph Male 67 29 3 Compact118 George Male 71 12 2 Compact201 Peter Male 54 8 2 Small

family202 Carl Male 72 30 2 Compact203 Walter Male 78 18 2 Executive204 Fred Male 70 33 2 Compact205 William Male 67 14 1 Compact206 Frank Male 60 19 2 Luxury207 Leonard Male 66 21 2 Executive208 Robert Male 70 26 2 Executive301 Dennis Male 71 21 2 Compact302 Harold Male 54 14 2 Compact303 Larry Male 59 17 5 Compact304 Ralph Male 58 11 1 Small

family305 Henry Male 55 21 2 Small

family306 Richard Male 56 43 2 Compact307 Patrick Male 50 18 4 Compact308 Eric Male 58 9 2 Compact309 Christopher Male 56 10 2 Small

family310 Ronald Male 63 24 2 Compact311 Kenneth Male 57 32 2 Compact312 James Male 41 12 2 Executive401 Brian Male 48 25 3 Compact402 Gerald Male 73 11 2 Small

family403 Betty Female 31 9 2 Small

family

Author's personal copy

412 G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413

Appendix A (Continued )

Object Nr Name Gender Age Monthssinceswitch

Householdsize

Last carmodel

404 Mary Female 55 18 2 Luxury405 Jeffrey Male 61 18 2 Small

family

References

Baumgartner, Martin (2008), “Interview on Regain Management,” Call CenterAktuell, 1 (5), 37.

Bowlby, John. (1973), Attachment and Loss, New York: Basic Books.Brady, Michael K., J.Joseph. Cronin Jr., Gavin L. Fox and Michelle L. Roehm

(2008), “Strategies to Offset Performance Failures: The Role of BrandEquity,” Journal of Retailing, 84 (2), 151–64.

Busboom, Amy L., Dawn M. Collins, Michelle D. Givertz and Lauren A. Levin(2002), “Can We Still Be Friends? Resources and Barriers to FriendshipQuality after Romantic Relationship Dissolution,” Personal Relationships,9, 215–23.

Calinski, T. and J. Harabasz (1974), “A Dendrite Method for Cluster Analysis,”Communications in Statistics – Theory and Methods, 3 (1), 1–27.

Clarke-Stewart, Alison and Cornelia Brentano (2006), Divorce: Causes andConsequences, New Haven, CT: Yale University Press.

Coulter, Robin A. and Mark Ligas (2000), “The Long-Good-Bye: The Dissolu-tion of Customer-Service Provide Relationships,” Psychology & Marketing,17 (8), 669–95.

Dant, Rajiv P. and Gregory T. Gundlach (1998), “The Challenge of Auton-omy and Dependence in Franchised Channels of Distribution,” Journal ofBusiness Venturing, 14, 35–67.

Duda, Richard O. and Peter E. Hart (1973), Pattern Classification and SceneAnalysis, New York: Wiley.

Duhachek, Adam (2005), “Coping: A Multidimensional, Hierarchical Frame-work of Responses to Stressful Consumption Episodes,” Journal ofConsumer Research, 32 (1), 41–53.

Dwyer, F. Robert, Paul H. Schurr and Sejo Oh (1987), “Developing Buyer–SellerRelationships,” Journal of Marketing, 51 (2), 11–27.

Folkman, Susan and Richard S. Lazarus (1988), “The Relationship BetweenCoping and Emotion: Implications for Theory and Research,” Social ScienceMedicine, 26, 309–17.

Fournier, Susan (1998), “Consumers and Their Brands: Developing RelationshipTheory in Consumer Research,” Journal of Consumer Research, 24, 343–74.March.

Gifi, Albert (1990), Nonlinear Multivariate Analysis, Chichester, UK:Wiley.

Glaser, Barney and Strauss Strauss (1967), The Discovery of Grounded Theory:Strategies for Qualitative Research, New York: Aldine de Gruyter.

Grégoire, Yany and Robert J. Fisher (2008), “Customer Betrayal and Retaliation:When Your Best Customers Become Your Worst Enemies,” Journal of theAcademy of Marketing Science, 36, 247–61.

Grewal, Dhruv, Julie Baker, Michael Levy and Glenn B. Voss (2003), “TheEffects of Wait Expectations and Store Atmosphere Evaluations on Patron-age Intentions in Service-Intensive Retail Stores,” Journal of Retailing, 79,259–68.

Gwinner, Kevin P., Dwayne D. Gremler and Mary Jo Bitner (1998), “RelationalBenefits in Services Industries: The Customer’s Perspective,” Journal of theAcademy of Marketing Science, 26 (2), 101–14.

Hair, Joseph F. Jr., William C. Black, Barry J. Babin, Rolph E. Anderson andRonald L. Tatham (2006), Multivariate Data Analysis, 6th edition UpperSaddle River: Prentice Hall.

Hemetsberger, Andrea, Christine M.T. Kittinger-Rosanelli and Sandra Fried-mann (2009), “‘Bye Bye Love’ – Why Devoted Consumers Break up withTheir Brands,” in Advances in Consumer Research North American Confer-ence Proceedings, 36, 430–7.

Hess, Ronald L., Shankar Ganesan and Noreen M. Klein (2007),“Interactional Service Failures in Pseudorelationship: The Roleof Organizational Attribution,” Journal of Retailing, 83 (1), 79–95.

Hirji, Karim F., Cyrus R. Mehta and Nitin R. Patel (1987), “Computing Distri-butions for Exact Logistic Regression,” Journal of the American StatisticsAssociation, 82, 1110–7.

Howell, Elizabeth F. (2008), The Dissociative Mind, London:Routledge.

Keaveney, Susan M. (1995), “Customer Switching Behavior in ServiceIndustries: An Exploratory Study,” Journal of Marketing, 59 (2),71–83.

Keaveney, Susan M. and Madhavan Parthasarathy (2001), “Customer SwitchingBehavior in Online Services: An Exploratory Study of the Role of SelectedAttitudinal, Behavioral, and Demographic Factors,” Journal of the Academyof Marketing Science, 29, 374–90 (Fall).

Kressmann, Frank, M. Joseph Sirgy, Andreas Herrmann, Frank Huber, StephanieHuber and Dong-Jin Lee (2006), “Direct and Indirect Effects of Self-Image Congruence on Brand Loyalty,” Journal of Business Research, 59(9), 955–64.

Linting, Marielle, J.J. Meulman, P.J.F. Groenen and J.J. van der Koojj (2007),“Nonlinear Principal Components Analysis: Introduction and Application,”Psychological Methods, 12 (3), 336–58.

Masheter, Carol (1991), “Postdivorce Relationships between Ex-Spouses: TheRoles of Attachment and Interpersonal Conflict,” Journal of Marriage andthe Family, 53 (1), 103–10.

(1997), “Healthy and Unhealthy Friendship and Hos-tility between Ex-Spouses,” Journal of Marriage and the Family, 59,463–75.

Mattilla, Anna S. and Paul G. Patterson (2004), “The Impact of Culture onConsumers’ Perceptions of Service Recovery Efforts,” Journal of Retailing,80, 196–20.

Mehta, Patel (1995), “Exact Logistic Regression: Theory and Examples,” Statis-tics in Medicine, 14 (19), 2143–60.

Miller, Dale (2001), “Disrespect and the Experience of Injustice,” Annual Reviewof Psychology, 52, 527–53.

Milligan, Glenn W. and Martha C. Cooper (1985), “An Examination of Proce-dures for Determining the Number of Clusters in a Data Set,” Psychometrika,50 (2), 159–7.

Mittal, Vikas, John W. Huppertz and Adwait Khare (2008), “Customer Com-plaining: The Role of Tie Strength and Information Control,” Journal ofRetailing, 84 (2), 195–204.

Olson, Wendy (2004), “Triangulation in Social Research: Qualitative and Quan-titative Methods Can Really be Mixed,” in Developments in Sociology,Holborn M. ed. Ormskirk: Causeway Press.

Orbuch, Tern L. (1997), “People’s Accounts Count: The Sociology of Accounts,”Annual Review of Sociology, 23 (1), 455–78.

Pan, Ye and George M. Zinkhan (2006), “Determinants of Retail Patron-age: A Meta-Analytical Perspective,” Journal of Retailing, 82 (3),229–43.

Parment, Anders (2008), “Distribution Strategies for Volume and PremiumBrands in highly Competitive Consumer Markets,” Journal of Retailing,15, 250–65.

Perreault, William D. Jr. and Forrest W. Young (1980), “Alternating LeastSquares Optimal Scaling: Analysis of Nonmetric Data in MarketingResearch,” Journal of Marketing Research, 17, 1–3.

Seetharaman, P.B. and Hai Che (2009), “Price Competition in Mar-kets with Consumer Variety Seeking,” Marketing Science, 28 (3),516–25.

Strauss, Anselm and Juliet Corbin (1998), Basics of Qualitative Research. Tech-niques and Procedures for Developing Grounded Theory, Newbury Park,CA: Sage.

Stauss, Bernd and Christian Friege (1999), “Regaining Service Customers, Costsand Benefits of Regain Management,” Journal of Service Research, 1 (4),347–61.

Tajfel, H. and J.C. Turner (1986), “The Social Identity Theory of Inter-GroupBehavior,” in Psychology of Intergroup Relations, Worchel S. and Austin L.W., eds. Chicago: Nelson-Hall.

Author's personal copy

G. Odekerken-Schröder et al. / Journal of Retailing 86 (4, 2010) 400–413 413

Thomas, Jacquelyn S., Robert C. Blattberg and Edward J. Fox (2004), “Recap-turing Lost Customers,” Journal of Marketing Research, 41, 31–45.

Tokman, Mert, Lenita M. Davis and Katherine N. Lemon (2007), “The WOWFactor: Creating Value through Win-Back Offers to Reacquire Lost Cus-tomers,” Journal of Retailing, 83 (1), 47–64.

von Wangenheim, Florian (2005), “Postswitching Negative Word of Mouth,”Journal of Service Research, 8 (1), 67–78.

Weiner, Bernard (1985), “An Attributional Theory of Achieve-ment Motivation and Emotion,” Psychological Review, 92 (4),548–73.

Weiss, R.S. (1975), Marital Separation, New York: Basic Books.Weitzmann, E.A. (2000), “Software and Qualitative Research,” in Handbook of

Qualitative Research, Denzin N. and Lincoln Y., eds. Thousand Oaks, CA:Sage, 803–20.