A critical review of techniques forclassifying quality attributes in

the Kano modelJosip Mikulic and Darko Prebezac

Department of Tourism, Faculty of Economics and Business,University of Zagreb, Zagreb, Croatia

Abstract

Purpose – The purpose of this paper is: to review the most commonly used approaches to theclassification of quality attributes according to the Kano model; to identify the theoretical/practicalstrengths and weaknesses of these techniques; and to provide guidance for future research andmanagerial practice in this area.

Design/methodology/approach – Based on an extensive review of the literature on the Kanomodel and the relevant marketing/management literature, five approaches (Kano’s method;“penalty-reward contrast analysis”; “importance grid”; qualitative data methods; and “directclassification”) are evaluated in terms of their validity and reliability for categorising attributes in theKano model. Several illustrative examples provide empirical evidence for the theoretical argumentsadvanced in the study.

Findings – The Kano questionnaire and the direct-classification method are the only approaches thatare capable of classifying Kano attributes in the design stage of a product/service. Penalty-rewardcontrast analysis (PRCA) is useful for assessing the impact of product/service attributes on overallsatisfaction with a product/service, but its applicability to the classification of Kano attributes isquestionable. The importance grid (IG) is not recommended for use with the Kano model. The criticalincident technique and the analysis of complaints/compliments are valid for the Kano model, but havequestionable reliability.

Originality/value – The study makes some important points about accurate semantic terminologyin describing issues related to the Kano model. In particular, researchers should be aware that anattractive quality element (must-be quality element, respectively) might in fact be a dissatisfier(satisfier, respectively), due to significant conceptual differences between performance in terms of theKano model (i.e. objective performance) and subjective performance perceptions.

Keywords Quality management, Classification, Critical incident technique, Marketing models,Customer satisfaction

Paper type General review

1. IntroductionSince its introduction in the 1980s, Kano’s model of attractive quality (Kano et al., 1984)has become one of the most popular quality models among marketing/managementpractitioners and researchers in a wide range of industries. The Kano model abandonsa strictly linear view of the impact of product/service attribute-performance oncustomer satisfaction in favour of identifying the particular attributes that have thepotential to elicit customer satisfaction/delight and/or dissatisfaction/frustration.

As the Kano model has become more popular, various conceptual approaches havebeen suggested for classifying quality attributes in the model. Examples include:

. Kano’s method;

. the “penalty-reward contrast analysis”;

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/0960-4529.htm

MSQ21,1

46

Managing Service QualityVol. 21 No. 1, 2011pp. 46-66q Emerald Group Publishing Limited0960-4529DOI 10.1108/09604521111100243

. the “importance grid”;

. qualitative data methods; and

. the “direct classification” method.

Several authors have conducted empirical comparisons of two or more of theseapproaches (Matzler and Sauerwein, 2002; Fuchs and Weiermair, 2003; Busacca andPadula, 2005; Bartikowski and Llosa, 2004; Witell and Lofgren, 2007). However, theresults have been mixed and no consensus has emerged on the most appropriateapproach. The purpose of the present study is, therefore, to build on the findings ofthese empirical studies with a view to providing guidance to researchers and managersin choosing and applying an appropriate approach for their needs.

In pursuing this objective, the present study does not attempt to conduct acomprehensive empirical comparison of the various approaches to attributeclassification in Kano’s model because it is not possible to make comparativeassessments of validity and reliability of particular methods/techniques unless it isknown that at least one of them provides reliable and valid results (and unless it isknown which provides such “true” results). Moreover, it would be very difficult to testall the approaches using the same sample of respondents and identical product/servicesettings.

The evaluation of the various approaches in this study therefore proceeds on thebasis of a comparative critique of their apparent strengths and weaknesses. Inparticular, the following three issues are addressed:

(1) the validity and reliability of the approaches (in terms of their theoreticalfoundations);

(2) the informational value of the results provided by each approach (in terms ofwhat exactly is being assessed); and

(3) the technical strengths and weaknesses of the approaches (in terms of theirmethodologies).

The findings of this study have practical application in assisting researchers to:. choose the most appropriate approach in a particular research setting;. interpret the results obtained through the various approaches; and. apply the various approaches in a technically appropriate manner.

The study also addresses several theoretical problems by identifying a number ofterminological and conceptual issues that require resolution if theory development inthis area is to be advanced.

The remainder of the paper is organised as follows. The first of five approaches(Kano’s method) is discussed in section 2. The “penalty-reward contrast analysis” isthen examined in section 3. The “importance grid” is assessed in section 4, andqualitative data methods are analysed in section 5. The last of the five approaches, the“direct classification method”, is discussed in section 6. A summary of the merits anddeficiencies of the main approaches is presented in section 7. The paper concludes witha discussion of the main findings and implications.

Classifyingquality attributes

47

2. Kano’s methodThe original creators of the Kano model provided their own approach to classifyingquality attributes using a structured questionnaire consisting of pairs of questions foreach attribute of a given product/service (Kano et al., 1984). One question in each pairasks about the consumer’s feelings in the case of fulfilment of an attribute (a so-called“functional question”) and the other question asks about feelings in the case ofnon-fulfilment of an attribute (“dysfunctional question”). The data are then analysedby means of a special evaluation table, which results in a categorisation of attributesfor each respondent (see Figure 1). Frequencies of single-respondent categorisationsare then used to provide the final classification of attributes.

Figure 1.The Kano method

MSQ21,1

48

Numerous studies have employed the Kano method, particularly in the field of totalquality management (TQM), in which Kano classifications have been used to adjustimprovement ratios in quality-function deployment (Franceschini and Terzago, 1998;Matzler and Hinterhuber, 1998; Tan et al., 1999; Shen et al., 2000; Tan and Shen, 2000; Tanand Pawitra, 2001; Pawitra and Tan, 2003; Lee et al., 2008; Baki et al., 2009). To improvethe original methodology, several authors have modified the wording of the questionsand/or the answers (Schvaneveldt et al., 1991; Tontini, 2000; Chen and Lee, 2009). Othershave modified the subcategories of the original Kano categories (Shahin and Zairi, 2009).

The rationale of the Kano method is logically valid for assessment of customerfeelings regarding a particular product/service attribute if the fulfilment (ornon-fulfilment) of that attribute is clearly defined. However, a significant problemcan emerge if the definitions of “fulfilment” and “non-fulfilment” are imprecise. Forexample, some authors frequently refer to quality attributes as being “attractive” ifthey elicit customer satisfaction when delivered, but do not cause dissatisfaction whenthey are not delivered; however, the same authors sometimes refer to attributes asbeing “attractive” if they have a greater positive impact on customer satisfaction whentheir performance is high than their negative impact when their performance is low(Fuchs and Weiermair, 2004; Matzler et al., 2004a, b; Fuller and Matzler, 2008).Moreover, some authors even combine these definitions and describe quality attributesas “attractive” if they cause customer satisfaction when they “are present or havesufficient performance”, but do not cause dissatisfaction when “not present or theirperformance is insufficient” (Tontini and Silveira, 2007, p. 486). It is thus beingimplicitly assumed that consumers’ evaluations of a given attribute in terms ofprovision/non-provision are the same (or at least similar) to consumers’ evaluations ofthe same attribute in terms of high/low attribute-performance. Acceptance of thisassumption implies that the Kano method provides the same (or at least similar)categorisations of attributes regardless of whether “fulfilment” and “non-fulfilment”are defined in terms of presence or defined in terms of performance.

To test the veracity of this assumption, the present study conducted a smallindicative survey using the example of a mobile banking service (that is, aninternet-based banking service delivered by mobile phone) as a feature of a new bankaccount. Two sets of questions were asked in this indicative survey of 129undergraduate students attending a market research class. The first set of questionsreferred to the provision/non-provision of this feature:

. Q1a: functional: “When opening a new bank account, how would you feel if youwere provided with a mobile banking service?”

. Q2a: dysfunctional: “When opening a new bank account, how would you feel ifyou were not provided with a mobile banking service?”

The second set of questions referred to the performance/non-performance of thefeature:

. Q1b: functional: “When opening a new bank account, how would you feel if youwere provided with a mobile banking service that works very well?”

. Q2b: dysfunctional: “When opening a new bank account, how would you feel ifyou were provided with a mobile banking service that works very poorly?”

The results of this brief questionnaire are provided in Table I.

Classifyingquality attributes

49

It is apparent from this illustrative example that the classification of the attributevaried significantly – depending on how “fulfilment” was defined and operationalised.When “fulfilment” was understood in terms of provision of the attribute, the mobilebanking service was classified as an “attractive” element, whereas when it wasunderstood in terms of performance, it was classified as a “must-be” element. Giventhat t-tests showed that both classifications were statistically significant at the 10 percent level, this failure to categorise the attribute consistently raises questions about thereliability and validity of this first approach (the Kano method) as a means ofcategorising attributes in the Kano model.

Nevertheless, interpretation of these apparently divergent results should take atleast two factors into consideration. First, it is acknowledged that the second(performance-based) operationalisation of the notion of “fulfilment” involved leadingquestions about performance that were likely to have caused significant response bias.Second, it is acknowledged that the two sets of questions presented quite differentsituations to the respondents – in that the first implied that the proposed new attributewill perform well, whereas the second explicitly allowed for the possibility that it mightperform well or poorly. These different hypothetical situations are likely to elicit quitedifferent responses, and, consequently, different attribute-categorisations.

To overcome these difficulties, it should be recognised that the key issue thatdetermines the Kano category of an attribute is not the performance of that attribute;rather, it is actually the provision (or non-provision) of a more-or-less expected benefit.It should also be recognised that the Kano model is talking in terms of objectiveperformance (e.g. size, weight, remote distance of a remote-control), and that lowobjective performance (a heavy remote control), does not have to be perceived as lowperformance by customers at a given point in time (e.g. when remote-controls were firstintroduced)! Accordingly, using the Kano method with performance-basedoperationalisations of “fulfilment” and “non-fulfilment” could make assessments ofthe different Kano-categories unreliable. To further increase the reliability of attributecategorisations, Kano’s method should refer to the provision (or non-provision) of thebenefits to be expected through the provision of an attribute rather than the provisionof the attribute itself. Using the example of the mobile banking service, the questionsshould therefore be formulated as follows:

. Q1c: functional: “When opening a new bank account, how would you feel if youwere provided with the possibility of managing your bank transactions via yourmobile phone?”

. Q2c: dysfunctional: “When opening a new bank account, how would you feel ifyou were not provided with the possibility of managing your bank transactionsvia your mobile phone?”

Mobile banking service A M O R Q I Total Category Significance

Provision/non-provision 59 10 19 0 0 41 129 A 0.061High/low performance 22 49 36 0 0 22 129 M 0.016

Note: Based on different wording of questions: provision-based wording; performance-based wordingTable I.Attribute classifications

MSQ21,1

50

Questions expressed in this way would provide a valid and reliable tool for assessingcustomer feelings regarding the provision (or non-provision) of a given attribute andthe benefits provided through it. However, a disadvantage of the technique is that itdoes not provide insight into the potential of that attribute to influence overallcustomer satisfaction, as opposed to satisfaction with the attribute itself. For example,this technique might reveal that the mobile banking service has greater potential tocreate satisfaction than dissatisfaction (that is, it is an “attractive” quality attribute),but the technique does not reveal how important this attribute is in the customer’soverall judgment of the bank account, nor how it relates to other attributes in thisregard. This is a significant disadvantage of this technique because such informationcan be important for effective decision-making when designing and/or improvingproducts/services.

Nevertheless, Kano’s method retains certain advantages over other approaches.First, it does not require the customer to have had experience with the attributes thatare being classified. Second, in contrast to other available approaches (except the“direct classification approach”, which is discussed later), the Kano method has notechnical limitations regarding the number of attributes that can be analysed.

3. Penalty-reward contrast analysisAlthough the technique of “penalty-reward contrast analysis” (PRCA) was originallydeveloped by Brandt (1987) to identify value-enhancing attributes in the specificcontext of transportation services, the PRCA approach has subsequently become afrequently used technique for classifying Kano attributes in a wide range of industries(Matzler and Sauerwein, 2002; Ting and Chen, 2002; Fuchs and Weiermair, 2003;Bartikowski and Llosa, 2004; Conklin et al., 2004; Fuchs and Weiermair, 2004; Matzleret al., 2004a, b; Busacca and Padula, 2005; Fuller et al., 2006; Matzler and Renzl, 2007;Fuller and Matzler, 2008; Alegre and Garau, 2009; Mikulic and Prebezac, 2010).

The technique analyses the impact of very high and very low attribute performanceon overall satisfaction using regression analysis with two sets of dummy variables foreach attribute. One set is obtained by coding the lowest attribute performance as “1”,whereas all other ratings of attribute performance are coded as “0”. The second set isobtained by coding the highest attribute performance as “1”, whereas all other ratingsof attribute performance are coded as “0”. The two dummy sets are then regressedagainst a measure of overall satisfaction, resulting in two regression coefficients (RCs)for each attribute. The first RC quantifies an attribute’s penalty (that is, the attribute’snegative impact on overall satisfaction when performing very poorly), whereas thesecond RC quantifies an attribute’s reward (that is, the attribute’s positive impact onoverall satisfaction when performing very well). A comparison of the two RCs thenreveals whether an attribute has a positive or negative impact on overall satisfaction.Moreover, the total of the absolute value of the penalty RC and the reward RC can beutilised to obtain a measure of the range of attribute impact on overall satisfaction,thus indicating an attribute’s importance in explaining the customer’s overalljudgment of the product/service (Mikulic and Prebezac, 2008).

Most authors use multiple regression analysis (MRA) in PRCA, although Shapleyvalue regression has been used in one study (Conklin et al., 2004), and bivariateregressions have been used in another study (Ting and Chen, 2002). The use of MRA inPRCA means that the technique is usually applied to a smaller set of attributes than

Classifyingquality attributes

51

other techniques (such as the Kano method or the “importance grid”) because MRA haslimitations in analysing larger numbers of variables due to the problem ofmulticollinearity. To circumvent this problem, some authors factor analyse the databefore applying PRCA (Matzler et al., 2004b; Fuller et al., 2006; Fuller and Matzler,2008). However, this is less than optimal because it does not facilitate detailed analysisof individual attributes of products/services; moreover, factor analysis is notnecessarily a valid approach because the input data for PRCA are obtained fromcustomer-satisfaction surveys, which frequently treat attributes as formativeindicators, rather than correlated factors.

Ting and Chen (2002) sought to circumvent the multicollinearity-problem by using abivariate approach, which facilitates analysis of any desired number of attributes;however, bivariate regressions are limited because the impact of all other attributes isneglected when calculating a given attribute’s impact on overall satisfaction.

An alternative approach to the problem of multicollinearity was proposed byMikulic and Prebezac (2009), whose hierarchical attribute model kept the number ofregressors in an equation at a reasonable level by splitting up products/services intotheir main components, and each component into a smaller number of attributes. Theattributes were then regressed against the respective component-satisfaction scores toreveal asymmetric effects within each component, whereas component-satisfactionwas regressed against overall satisfaction with the product/service to revealasymmetries at the product/service level.

Another important issue to consider when using PRCA is that differences inexpectations among different customer segments can cause variations in perceptionsof attribute performance that do not accord with objective variations in theperformance of various attributes. As a consequence, Fuller and Matzler (2008) havesuggested that researchers should consider a segment-based approach wheninterpreting PRCA results.

A neglected technical issue in PRCA is the widespread use of standardised RCs inPRCA studies (Matzler and Sauerwein, 2002; Fuchs and Weiermair, 2003; Fuchs andWeiermair, 2004; Matzler et al., 2004a; Matzler and Renzl, 2007; Fuller and Matzler,2008; Mikulic and Prebezac, 2008). This is despite evidence that such variables skewthe information held in the original unstandardised variables – unless the dummyvariables have equal distributions of ones and zeros, which would imply they have thesame arithmetic means and standard deviations (Fox, 1997). To meet this condition,performance data for each attribute would have to cover the same frequencies of lowestand highest performance-scores. However, because this condition is rarely met,standardised and unstandardised RCs are likely to have quite different implications.An illustrative example is provided in Table II. The data were gathered from a surveyon airline customer satisfaction ðn ¼ 997Þ: Attribute performance and overallsatisfaction were measured on seven-point Likert scales. Only lowest and highest APscores were coded into dummy variables (that is, AP ¼ 1 and 7).

The results reveal that only one attribute shows the same direction ofimpact-asymmetry for both types of RCs (flight experience). Given these divergentresults, it is strongly recommended that standardised dummy RCs should not be usedin PRCA because they are likely to have misleading implications.

In summary, when PRCA is used properly it is a reliable approach for analysingasymmetries in the relationship between subjective perceptions of attribute

MSQ21,1

52

performance and overall satisfaction with a product/service. Compared with the Kanomethod, PRCA has advantages and disadvantages. Its advantage is that it can be usedto discriminate among attributes in terms of their relative importance in explaining acustomer’s overall judgment of a product/service; conversely, its disadvantage is that,unlike the Kano method, it cannot be used to analyse attributes that do not yet exist.

Nevertheless, the conceptual validity of PRCA as a Kano classification method isquestionable because PRCA analyses the relationship between attribute performanceand overall satisfaction, whereas the Kano model analyses the relationship betweenattribute performance and attribute satisfaction. Although it could be argued thatoverall satisfaction is a proxy measure of attribute satisfaction, doubts remain aboutthe conceptual validity of PRCA as a Kano classification tool – because the Kanomodel purports to describe objective attribute performance, whereas the PRCA relieson subjective perceptions of attribute performance. Although these can converge undercertain circumstances, it is clear that the Kano model and the PRCA do not assess thesame things. Indeed, it seems more appropriate to refer to the categories of attributesunder PRCA as “satisfiers”, “dissatisfiers”, and “hybrids”, rather than using theclassifications of the Kano model. This distinction is suggested because the conceptualdifferences between objective attribute performance and subjective attributeperformance mean that an “attractive” quality attribute under the Kano modelmight, in theory, be a “dissatisfier” under PRCA, and a “must be” quality attributeunder the Kano model might, in theory, be a “satisfier” under PRCA.

4. “Importance grid”The “importance grid” (IG), which was developed by an IBM consultant as ananalytical tool for classifying quality attributes in the Kano model, was first cited in thescholarly literature by Vavra (1997). Since then it has been applied to a variety ofproducts (Matzler and Hinterhuber, 1998; Yang, 2005; Riviere et al., 2006) and services(Martensen and Gronholdt, 2001; Fuchs, 2002; Fuchs and Weiermair, 2003, 2004;Matzler et al., 2003; Bartikowski and Llosa, 2004; Busacca and Padula, 2005).

The IG compares explicit and implicit “attribute importance” (AI) to classifyattributes. Explicit AI is obtained directly from the customer (through direct ratings or

Unstandardised coefficients Standardised coefficients

Attribute Penalty Reward

Direction ofimpact 2asymmetry Penalty Reward

Direction ofimpact 2asymmetry

Flight offer 20.431 * 0.344 * * Negative 20.035 * 0.062 * * Positive

Ticket purchaseexperience 20.827 * 0.675 * * * Negative 20.044 * 0.128 * * * Positive

Flight experience 20.510 * 1.158 * * * Positive 20.031 * 0.220 * * * Positive

Price 21.186 * * * 0.710 * * * Negative 20.096 * * * 0.128 * * * Positive

Ontime performance 20.721 * * 0.722 * * * Linear 20.053 * * 0.137 * * * Positive

Notes: *p , 0.05; * *p , 0.01; * * *p , 0.001; R 2 ¼ 5.41; Adjusted R 2 ¼ 0.532; Dependentvariable ¼ overall satisfaction

Table II.Unstandardised vs

standardised regressioncoefficients in PRCA

Classifyingquality attributes

53

rankings), whereas implicit AI is statistically derived by regressing or correlatingattribute performance against a global performance measure (such as overallsatisfaction). Most authors have used standardised beta coefficients from MRA (Fuchs,2002; Matzler and Sauerwein, 2002; Matzler et al., 2002; Busacca and Padula, 2005;Peters, 2005) or partial correlation coefficients (Matzler et al., 2002, 2003; Fuchs andWeiermair, 2003, 2004; Bartikowski and Llosa, 2004), although some authors haveutilised Pearson- or Spearman rank-order correlation coefficients (Matzler et al., 2002).

To classify the attributes, the measures of explicit AI and implicit AI are used toconstruct a two-dimensional grid, which is divided into four quadrants using grandmeans of implicit and explicit AI scores as thresholds. By noting the positions ofattributes on this grid, the analysis purports to reveal the various Kano factors asfollows:

(1) “must-be” attributes have high (above average) explicit AI, but low(below-average) implicit AI;

(2) “attractive” attributes have low explicit AI, but high implicit AI;

(3) “one-dimensional” attributes with high importance have high explicit AI andhigh implicit AI; and

(4) “one-dimensional” attributes with low importance have low explicit AI and lowimplicit AI.

Several authors have questioned the validity of the IG technique because there is notheory that explains why the different Kano factors can be identified by comparingexplicit AI with implicit AI (Matzler and Sauerwein, 2002; Fuchs and Weiermair, 2003;Busacca and Padula, 2005). In particular, the implicit assumption that the differentKano categories can be identified by comparing the relative grid positions of attributesmeans that the categorisation of any given attribute is dependent on all the otherattributes as reference points. This has two logical consequences. First, the IG willalways yield a classification of attributes into all of the various Kano categories(“must-be”, “attractive”, and “one-dimensional”) whenever there are any differences inexplicit and implicit AI (which is usually the case); however, it is doubtful that every setof analysed attributes always covers every Kano category. Second, a given attribute’sposition on the grid (and hence its Kano categorisation) can change if the set ofanalysed attributes is modified; however, according to the Kano model, thecategorisation of a given attribute should be objectively defined and consistentacross any set of attributes.

Figures 2 and 3 illustrate how the choice of analysed attributes can influence IGresults. Both IG “A” (Figure 2) and IG “B” (Figure 3) are based on the same data onairline customer satisfaction, with the only difference being that IG “B” encompassesone more attribute than IG “A”.

It is apparent that the “ticket purchase experience” is classified as a“one-dimensional” attribute in IG “A”, but as a “must-be” attribute in IG “B”.Moreover, “price” is an “attractive” attribute in IG “A”, but a “one-dimensional”attribute in IG “B”. This example demonstrates that the IG is an unreliable techniquefor classifying attributes into predefined Kano categories.

Despite these problems, the IG technique does have some merit. First, explicit AI canbe regarded as an indicator of expectations with regard to attributes (Lambert and

MSQ21,1

54

Lewis, 1990). This is in accordance with the spirit of the Kano model, which holds that“attractive” attributes are less expected, whereas “one-dimensional” and “must-be”attributes are more expected. Second, implicit AI is an indicator of an attribute’s impacton customer satisfaction, which would also seem to be in the spirit of the Kano model,which holds that “attractive” attributes and “one-dimensional” attributes have asignificant impact on satisfaction, whereas “must-be” attributes have less impact onsatisfaction. However, it should also be noted that Kano’s model holds that“one-dimensional” attributes and “must-be” elements also have a significant impact ondissatisfaction, which is a problem for the IG technique because implicit AI indicates anattribute’s potential to impact overall satisfaction (ranging from very dissatisfied to verysatisfied). As a consequence, because implicit AI does not reveal whether the impact of agiven attribute is stronger towards satisfaction or dissatisfaction, a measure thatindicates the asymmetry of impact is required. Such a measure has been introduced byMikulic and Prebezac (2008); however, with regard to the applicability of the IG as aKano assessment tool, this does not address the problem of how objective attributeperformance is to be reliably operationalised and the problem that the techniqueanalyses the relationship between attribute performance and overall satisfaction, ratherthan the relationship between attribute performance and attribute satisfaction.

In summary, the IG cannot be regarded as a reliable approach for assessingattributes into absolute categories in the Kano model. However, as Bartikowski andLlosa (2004) have contended, it could serve as an advanced version of theimportance-performance analysis (Martilla and James, 1977). Because the IG uses both

Figure 2.Importance-grid “A”

Classifyingquality attributes

55

direct and statistically derived measures of AI, which are the two most frequently usedAI measures that do not necessarily converge (Wiley et al., 1977; Barlas, 2003), acombination of these two dimensions of importance might result in greater reliabilityof relative attribute-prioritisations in improvement strategies (Mikulic and Prebezac,2009).

5. Qualitative data methodsThe “two-factor theory” of work satisfaction/work dissatisfaction and the criticalincident technique (CIT) (Herzberg et al., 1959) have been widely adopted by authors toexplore the notion that there are attributes that are primarily satisfying and attributesthat are primarily dissatisfying in various products (Hausknecht, 1988; Maddox, 1981;Swan and Combs, 1976) and services (Bitner et al., 1990; Johnston and Silvestro, 1990;Silvestro and Johnston, 1990; Smith et al., 1992; Stauss and Hentschel, 1992; Johnston,1995). In a similar vein, other authors have applied the technique of “analysis ofcomplaints and compliments” (ACC) for the same purpose (Cadotte and Turgeon, 1988;Oliver, 1997; Friman and Edvardsson, 2003). Although these various studies did notrefer directly to the Kano model, several authors refer to both CIT and ACC asqualitative data-based approaches for application in the Kano model (Matzler et al.,2002; Matzler and Sauerwein, 2002).

The basic assumption in applying CIT and ACC to the Kano model is that qualityattributes can be categorised by comparing how frequently customers mention anattribute in a positive context or a negative context. In summary:

Figure 3.Importance-grid “B”

MSQ21,1

56

. If the number of negative incidents/complaints significantly exceeds the numberof positive incidents/compliments, the attribute is categorised as a “must-be”attribute.

. If the number of positive incidents/compliments significantly exceeds thenumber of negative incidents/complaints, the attribute is categorised as an“attractive” attribute.

. If the number of positive incidents/compliments and the number of negativeincidents/complaints are approximately equal, the attribute is categorised as a“one-dimensional” attribute.

The theoretical assumptions in applying these techniques to the Kano model appear tobe appropriate. In essence, the theory is that “must-be” attributes prevail in negativecontexts because they have greater potential to cause dissatisfaction than satisfaction,whereas “attractive” elements prevail in positive contexts because they have a greaterpotential to cause satisfaction than dissatisfaction. However, in practice, a significantlimitation of these techniques is that they do not allow researchers to determine whichattributes are to be categorised because the attributes for consideration emerge onlyfrom the customers’ comments.

It is noteworthy that CIT and ACC are the only approaches under consideration in thepresent study that externally validate the assumptions of the Kano model – that is, theybegin with delight/frustration and analyse what caused it, whereas the Kano methodassumes fulfilment/non-fulfilment and analyses the consequent customer feelings(satisfaction or dissatisfaction). However, as noted above, it is theoretically possible that“attractive” attributes are “dissatisfiers” that cause frustration and that “must-be”attributes are satisfiers that cause delight. Thus, in order to make inferences about thereliability of these techniques for categorising attributes in the Kano model it would beimportant to know whether delight/frustration is primarily a result of extraordinaryperformance/failure of a more-or-less expected attribute, or whether it is a result oflow/high expectedness of a provided benefit that is a necessary condition for an attributeto be an “attractive”/”must-be” attribute. Moreover, as Matzler et al. (2002) have noted,the fact that some attributes predominate in dissatisfying incidents/complaints, whereasothers predominate in satisfying incidents/compliments, might be due to patterns ofattribute performance in a particular research setting – in other words, the product orservice under analysis might generally perform very well in some areas and very poorlyin others. In these circumstances, such analyses might therefore be indicative, but inorder to categorise the attributes reliably, the data would need to cover both very positiveand very negative customer experiences with each attribute that emerges from theanalysis. However, this might not always be achieved in case-based studies; indeed,given the methodological rationale of the techniques, it is usually impossible to ascertainwhether there were positive customer experiences that did not cause compliments (ornegative experiences that did not cause complaints) – although reliability certainlyincreases when studies are based on multiple cases.

In summary, the reliability of both the CIT and ACC to classify attributes in theKano model remains questionable. Nevertheless, the managerial value of the resultsremains high if the objective of the research is to identify active and/or salient sourcesof satisfaction (“satisfiers”) and dissatisfaction (“dissatisfiers”) in a particularproduct/service setting.

Classifyingquality attributes

57

6. Direct classificationA simple method of direct classification of attributes according to the Kano model wasproposed by Emery and Tian (2002), who suggested that the theory of attractive qualityshould be explained to respondents, who would then themselves classify attributes intothe various Kano categories. A similar two-step direct-classification approach, termedthe “force-choice classification method”, was suggested by Shen et al. (2000). As a firststep, respondents were asked to classify product/service attributes into the three mainKano categories (“attractive”, “must-be”, and “one-dimensional”), followed by a secondstep in which respondents refined the classification by sub-categorising attributes into“somewhat attractive/basic”, “moderately attractive/basic”, and “very attractive/basic”quality attributes.

Such direct classification is the only approach that does not rely on an indirectassessment of the various Kano categories on the basis of implicit assumptions.Moreover, reliability problems can occur only if the researcher does not explain thetheory well and/or the respondents do not understand it well. As with the firsttechnique examined in the present study (the Kano method), the advantages of directclassification are:

. that it can be used to classify both existing and non-existing attributes; and

. that it has no technical limitations regarding the number of attributes that can beanalysed.

However, the practical disadvantages of the direct-classification approach are thatexplaining the theory to respondents can be time-consuming and it can be difficult forsome respondents to understand it.

In summary, the disadvantages of the approach mean that the applicability of directclassification in typical customer-satisfaction surveys is limited, and reliabilityproblems can occur. Given that the Kano method does not require special training foreither interviewers or respondents, it would appear to be preferable to thedirect-classification method.

7. Summary of findingsTable III provides a summary of the approaches in terms of:

. what is being assessed in each;

. the validity of each as a means of categorising attributes in the Kano model; and

. the reliability of each.

8. Conclusions and implications8.1 Implications for researchersA key finding of this study is that the reliability and validity problems of most of theapproaches to attribute classification in the Kano model derive from differences between:

. the concept of objective attribute-performance (AP), as described in the Kanomodel; and

. the concept of subjective AP-perception, which is commonly used in attempts tooperationalise the Kano model.

Several important implications emerge from this conceptual difference.

MSQ21,1

58

First, given that the temporal dimension is firmly incorporated in the Kano model,AP-perceptions of the same levels of objective AP can differ significantly at variouspoints in time. For example, a large and bulky remote control for a television (havinglow objective AP according to the Kano model) might not have been perceived asperforming poorly in the 1950s when remote controls were first introduced; however,the same remote control would be likely to yield low AP-ratings in a consumer surveyconducted today. It is thus apparent that objective AP and subjectively perceived APare not equivalent concepts – because objective AP is free from chronologicalinfluences, whereas subjective AP-perceptions are strongly time-dependent (as aconsequence of changes in customer expectations, familiarity with a product/service,product/service innovations, and so on).

Approach What is being assessed?

Validity forassessing theKano model

Reliability for assessingthe Kano model

Kano method (provision/non-provision-basedmode)

Asymmetries in customerfeelings based onhypothesised provision/non-provision ofcustomer benefits/values

Yes High

Kano method (high/lowperformance-based mode)

Asymmetries in customerfeelings based onhypothesised high/lowlevels of attributeperformance

Yes Low

Penalty-reward contrastanalysis

Asymmetric effects inoverall satisfaction with aproduct/service based onperceived high/lowattribute performance

Questionable –

Importance grid Attribute importanceaccording to two differentdimensions of importance

No –

Critical incidenttechnique/analysis ofcomplaints andcompliments

Active sources ofsatisfaction anddissatisfaction (case-based studies)Salient sources ofsatisfaction anddissatisfaction (multi-case studies)

Yes Questionable (depends on(non-assessable) qualityof attribute performancedata)

Direct classification The belonging ofattributes to predefinedcategories (e.g. thedifferent attribute-categories of the Kanomodel)

Yes Questionable (depends onquality of interviewers/respondents)

Table III.Summary of findings

Classifyingquality attributes

59

Second, although the range of possible customer reactions to obtained objective APis pre-determined by an attribute’s degree of evolution according to the Kano model(that is, by its Kano category), the range of possible customer reactions to perceived APis in no way determined by an attribute’s Kano category. For example, although aremote control was an attractive attribute of a television in 1983 (Kano, 2001), there isno reason to assume that a consumer in those times might not have been verydissatisfied with the remote control (perhaps because it malfunctioned), or that aconsumer in 2010 might not be very satisfied (perhaps because it has an appealingdesign) – despite it being a “must-be” element. It is thus apparent that “attractive”attributes can in fact be “dissatisfiers” (that is, they cause greater dissatisfaction thansatisfaction), and that “must-be” elements can be “satisfiers”. It is therefore importantthat these terms not be used synonymously (which is, unfortunately, very common inthe extant literature).

Third, it is apparent from the above example that an “attractive” attribute can, infact, cause dissatisfaction because its AP is perceived to be low. As a consequence,certain definitions of “attractive” attributes in the literature that do ostensibly apply tothe Kano model (which speaks in terms of objective AP) are not necessarily applicableto the real world (which is perceived subjectively). Examples of such definitionsinclude:

. “increase customer satisfaction if delivered” (Fuchs and Weiermair, 2004, p. 217);

. do not cause dissatisfaction when “not fulfilled” (Nilsson-Witell and Fundin,2005, p. 157); and

. “when not present or their performance is insufficient” (Tontini and Silveira,2007, p. 486).

Similar comments hold for common definitions of “must-be” attributes and“one-dimensional” attributes. As a consequence, attempts to operationalise the Kanomodel by analysing how subjective AP-perceptions are related to resulting customerreactions (satisfaction/dissatisfaction) cannot be considered reliable.

Fourth, analyses of how objective AP is related to resulting customer reactionswould be reliable and theoretically valid for categorisation in the Kano model.However, because objective AP is typically a vague concept that encompassesseveral characteristics of a feature (for example, size of a remote control, its weight,power consumption, remote distance, and so on), it might be impossible to measurethe AP of a feature as a whole objectively – because such measures do not exist! Inthis regard, the Kano model seems to have a conceptual flaw – because theobjective AP of a feature can often be described only in terms of the objective AP ofthe feature’s characteristics. This, however, shifts the focus to a lower level ofabstraction which makes it possible to assess only whether the characteristics of thefocal feature are “attractive”, “one-dimensional”, or “must-be” attributes. It is thusapparent that attempts to operationalise the Kano model by analysing how providedobjective AP is related to resulting customer reactions can become an impossibletask.

Finally, with regard to the relationship between objective AP and the various Kanocategories, it should be recognised that the level of objective AP does not determinethe Kano category of a feature, but only the degree of customersatisfaction/dissatisfaction which ranges between different boundaries for different

MSQ21,1

60

Kano categories. Moreover, with the evolution of a feature towards a “must-be”attribute, the level of attribute satisfaction associated with a particular level ofobjective AP becomes lower for all levels of objective AP. In fact, it is theprovision/non-provision of a more-or-less expected and more-or-less useful customerbenefit that determines the feature’s Kano category (for example, in the 1950s, thiswas the provision of a not-expected feature that enabled the viewer to change thechannel without having to touch the television). Accordingly, in operationalising theKano model, it is strongly recommended that use be made of aprovision/non-provision-based mode of attribute fulfilment/non-fulfilment. Given therationale of the Kano model, such a mode is not only valid, but also very reliable –due to the fact that its manifestations can easily be operationalised in an unequivocaland objective way (that is, provided/not-provided).

8.2 Implications for managersManagers have a strong interest in identifying product/service features that are keydrivers of customer satisfaction and dissatisfaction. In this regard, classifyingproduct/service features according to the Kano model provides managers withvaluable information that can be used in designing new products/services or inimproving existing ones. However, managers should be aware that most of thecommonly used approaches for classifying attributes in the Kano model have thepotential to provide quite different categorisations of product/service attributes.Quality managers should therefore regard the various approaches as complementary,rather than as alternative approaches for the same purpose.

The study has shown that the Kano method is particularly useful in the designstage of a product/service because it facilitates categorisation of both existing andnon-existing product/service attributes according to their potential to elicitsatisfaction/delight and dissatisfaction/frustration. However, when using the Kanomethod, it is strongly recommended that a reasonable level of performance of theanalysed attributes should be presumed, and that the customers’ feelings should beassessed in the case of provision and non-provision of the benefits provided throughthe attributes under consideration. A shortcoming of the Kano method, however, is thatit does not reveal the relative importance of various attributes in the customer’s overallevaluation of the product/service, and/or how the analysed attributes relate to eachother in this regard.

To obtain such information, a penalty-reward contrast analysis (PRCA) can provideinsight. Although the validity of PRCA for categorisation of attributes in the Kanomodel is questionable, it does have considerable managerial value. For example, thePRCA quantifies penalties for low performance and rewards for high performance interms of changes in overall satisfaction with a product/service. Moreover, penalties andrewards are comparable across attributes, and they can be totalled to obtain a measureof an attribute’s impact on overall satisfaction with a product/service. However, thePRCA can be applied only to existing attributes.

Managers should also be aware that the importance-grid (IG) cannot be regarded asa valid technique for use with the Kano model. However, in combination with data onattribute-performance, the IG can serve as an advanced importance–performanceanalysis to obtain relative improvement-priorities of attributes based on two differentdimensions of attribute-importance.

Classifyingquality attributes

61

Managers should also be aware that qualitative-data methods, such as the criticalincident technique and the analysis of complaints and compliments, could providevaluable information about the most significant sources of customersatisfaction/dissatisfaction in a particular company/product/service setting.However, the reliability of these techniques for classifying attributes according tothe Kano model is questionable because it is impossible to discern whether theanalysed research setting covers both very positive and very negative customerexperiences for all attributes that emerge from the analysis (but which did not result incomplimenting/complaining behaviour).

Finally, direct classifications are certainly a valid approach to categorisingattributes in the Kano model. However, the reliability of these methods depends on howwell the interviewers explain the Kano model to respondents, and how wellrespondents understand the model. Given that the Kano method is also valid, but doesnot require special training of interviewers and respondents, it is recommended thatthe Kano method be preferred to direct classifications.

References

Alegre, J. and Garau, J. (2009), “The factor structure of tourist satisfaction at sun and sanddestinations”, Journal of Travel Research, Vol. 43 No. 3, pp. 226-37.

Baki, B., Basfirinci, C.S., Ar, I.M. and Cilingir, Z. (2009), “An application of integratingSERVQUAL and Kano’s model into QFD for logistics services: a case study from Turkey”,Asia Pacific Journal of Marketing and Logistics, Vol. 21 No. 1, pp. 106-26.

Barlas, S. (2003), “When choices give in to temptations: explaining disagreement amongimportance measures”, Organizational Behavior and Human Decision Processes, Vol. 91No. 2, pp. 310-21.

Bartikowski, B. and Llosa, S. (2004), “Customer satisfaction measurement: comparing fourmethods of attribute categorisations”, The Service Industries Journal, Vol. 24 No. 4,pp. 67-82.

Bitner, M.J., Booms, B.H. and Tetreault, M.S. (1990), “The service encounter: diagnosingfavorable and unfavorable incidents”, Journal of Marketing, Vol. 54, pp. 71-84.

Brandt, R.D. (1987), “A procedure for identifying value-enhancing service components usingcustomer satisfaction survey data”, in Surprenant, C.F. (Ed.), Add Value to Your Service:The Key to Success, AMA, Chicago, IL, pp. 61-4.

Busacca, B. and Padula, G. (2005), “Understanding the relationship between attributeperformance and overall satisfaction: theory, measurement and implications”, MarketingIntelligence & Planning, Vol. 23 No. 6, pp. 543-61.

Cadotte, E.R. and Turgeon, N. (1988), “Dissatisfiers and satisfiers: suggestions from consumercomplaints and compliments”, Journal of Consumer Satisfaction, Dissatisfaction andComplaining Behaviour, Vol. 1, pp. 74-9.

Chen, Y.K. and Lee, Y.C. (2009), “A new method to identify the category of the quality attribute”,Total Quality Management and Business Excellence, Vol. 20 No. 10, pp. 1139-52.

Conklin, M., Powaga, K. and Lipovetsky, S. (2004), “Customer satisfaction analysis: identificationof key drivers”, European Journal of Operational Research, Vol. 154 No. 3, pp. 819-27.

Emery, C.R. and Tian, R.G. (2002), “Schoolwork as products, professors as customers: a practicalteaching approach in business education”, Journal of Education for Business, Vol. 78 No. 2,pp. 97-102.

MSQ21,1

62

Fox, J. (1997), Applied Regression Analysis, Linear Models, and Related Methods, SagePublications, Thousand Oaks, CA.

Franceschini, F. and Terzago, M. (1998), “An application of quality function deployment toindustrial training courses”, International Journal of Quality & Reliability Management,Vol. 15 No. 7, pp. 753-68.

Friman, M. and Edvardsson, B. (2003), “A content analysis of complaints and compliments”,Managing Service Quality, Vol. 13 No. 1, pp. 20-6.

Fuchs, M. (2002), “Benchmarking indicator-systems and their potential for tracking guestsatisfaction”, Tourism, Vol. 50 No. 2, pp. 141-55.

Fuchs, M. and Weiermair, K. (2003), “New perspectives on satisfaction research in tourismdestinations”, Tourism Review, Vol. 58 No. 3, pp. 6-14.

Fuchs, M. and Weiermair, K. (2004), “Destination benchmarking: an indicator-system’s potentialfor exploring guest satisfaction”, Journal of Travel Research, Vol. 42, pp. 212-25.

Fuller, J. and Matzler, K. (2008), “Customer delight and market segmentation: an application ofthe three-factor theory of customer satisfaction on life style groups”, TourismManagement, Vol. 29, pp. 116-26.

Fuller, J., Matzler, K. and Faullant, R. (2006), “Asymmetric effects in customer satisfaction”,Annals of Tourism Research, Vol. 33 No. 4, pp. 1159-63.

Hausknecht, D. (1988), “Emotional measures of satisfaction/dissatisfaction”, Journal ofConsumer Satisfaction, Dissatisfaction and Complaining Behavior, Vol. 1, pp. 25-33.

Herzberg, F., Mausner, B. and Snyderman, B. (1959), The Motivation to Work, Wiley, New York,NY.

Johnston, R. (1995), “The determinants of service quality: satisfiers and dissatisfiers”,International Journal of Service Industry Management, Vol. 6 No. 5, pp. 53-71.

Johnston, R. and Silvestro, R. (1990), “The determinants of service quality – a customer-basedapproach”, Proceedings of the Decision Science Institute Conference, San Diego, CA,November.

Kano, N. (2001), “Life cycle and creation of attractive quality”, paper presented at the4th International QMOD Conference on Quality Management and OrganisationalDevelopment, University of Linkoping, Linkoping.

Kano, N., Seraku, N., Takahashi, F. and Tsuji, S. (1984), “Attractive quality and must-be quality”,Hinshitsu, The Journal of the Japanese Society for Quality Control, April, pp. 39-48.

Lambert, D.M. and Lewis, M.C. (1990), “A comparison of attribute importance and expectationscales for measuring service quality”, in Bearden, L. (Ed.), Enhancing KnowledgeDevelopment in Marketing, AMA, Chicago, IL.

Lee, Y., Sheu, L. and Tsou, Y. (2008), “Quality function deployment implementation based onfuzzy Kano model: an application in PLM system”, Computers & Industrial Engineering,Vol. 55 No. 1, pp. 48-63.

Maddox, R.N. (1981), “Two-factor theory and consumer satisfaction: replication and extension”,Journal of Consumer Research, Vol. 8, pp. 97-102.

Martensen, A. and Gronholdt, L. (2001), “Using employee satisfaction measurement to improvepeople management: an adaptation of Kano’s quality type”, Total Quality Management,Vol. 12 Nos 7/8, pp. 949-57.

Martilla, J.A. and James, J.C. (1977), “Importance-performance analysis”, Journal of Marketing,Vol. 41 No. 1, pp. 77-99.

Classifyingquality attributes

63

Matzler, K. and Hinterhuber, H.H. (1998), “How to make product development projects moresuccessful by integrating Kano’s model of customer satisfaction into quality functiondeployment”, Technovation, Vol. 18 No. 1, pp. 25-38.

Matzler, K. and Renzl, B. (2007), “Assessing asymmetric effects in the formation of employeesatisfaction”, Tourism Management, Vol. 28 No. 4, pp. 1093-103.

Matzler, K. and Sauerwein, E. (2002), “The factor structure of customer satisfaction: an empiricaltest of the importance grid and the penalty-reward-contrast analysis”, InternationalJournal of Service Industry Management, Vol. 13 No. 4, pp. 314-32.

Matzler, K., Fuchs, M. and Schubert, A.K. (2004a), “Employee satisfaction: does Kano’s modelapply?”, Total Quality Management and Business Excellence, Vol. 15 Nos 9/10, pp. 1179-98.

Matzler, K., Sauerwein, E. and Heischmidt, K.A. (2003), “Importance-performance analysisrevisited: the role of the factor structure of customer satisfaction”, The Service IndustriesJournal, Vol. 23 No. 2, pp. 112-29.

Matzler, K., Sauerwein, E. and Stark, C. (2002), “Methoden zur Identifikation von Basis-,Leistungs- und Begeisterungsfaktoren”, in Hinterhuber, H.H. and Matzler, K. (Eds),Kundenorientierte Unternehmensfuhrung: Kundenorientierung-Kundenzufriedenheit-Kundenbindung, Gabler, Wiesbaden, pp. 265-89.

Matzler, K., Bailom, F., Hinterhuber, H.H., Renzl, B. and Pichler, J. (2004b), “The asymmetricrelationship between attribute-level performance and overall customer satisfaction:a reconsideration of the importance-performance analysis”, Industrial MarketingManagement, Vol. 33, pp. 271-7.

Mikulic, J. and Prebezac, D. (2008), “Prioritizing improvement of service attributes using impactrange-performance analysis and impact-asymmetry analysis”, Managing Service Quality,Vol. 18 No. 6, pp. 559-76.

Mikulic, J. and Prebezac, D. (2009), “Developing service improvement strategies underconsideration of multicollinearity and asymmetries in loyalty intentions – a study from theairline industry”, Proceedings of the 11th International Conference QUIS – The ServiceConference, Wolfsburg, Germany, pp. 414-23.

Mikulic, J. and Prebezac, D. (2010), “Evaluating hotel animation programs at Mediterranean sunand sea resorts: an impact-asymmetry analysis”, Tourism Management, 25 June.

Nilsson-Witell, L. and Fundin, A. (2005), “Dynamics of service attributes: a test of Kano’s theoryof attractive quality”, International Journal of Service Industry Management, Vol. 16 No. 2,pp. 152-68.

Oliver, R.L. (1997), Satisfaction: A Behavioral Perspective on the Consumer, Irwin/McGraw-HillCompanies, New York, NY.

Pawitra, T.A. and Tan, K.C. (2003), “Tourist satisfaction in Singapore – a perspective fromIndonesian tourists”, Managing Service Quality, Vol. 13 No. 5, pp. 399-411.

Peters, M. (2005), “Entrepreneurial skills in leadership and human resource managementevaluated by apprentices in small tourism businesses”, Education þ Training, Vol. 47Nos 8/9, pp. 575-91.

Riviere, P., Monrozier, R., Rogeaux, M., Pages, J. and Saporta, G. (2006), “Adaptive preferencetarget: contribution of Kano’s model of satisfaction for an optimized preference analysisusing a sequential consumer test”, Food Quality and Preference, Vol. 17, pp. 572-81.

Schvaneveldt, S.J., Enkawa, T. and Miyakawa, M. (1991), “Consumer evaluation perspectives ofservice quality: evaluation factors and two-way model of quality”, Total QualityManagement, Vol. 2 No. 2, pp. 149-61.

MSQ21,1

64

Shahin, A. and Zairi, M. (2009), “Kano model: a dynamic approach for classifying and prioritisingrequirements of airline travellers with three case studies on international airlines”, TotalQuality Management & Business Excellence, Vol. 20 No. 9, pp. 1003-28.

Shen, X.X., Tan, K.C. and Xie, M. (2000), “An integrated approach to innovative productdevelopment using Kano’s model and QFD”, European Journal of InnovationManagement, Vol. 3 No. 2, pp. 91-9.

Silvestro, R. and Johnston, R. (1990), “The determinants of service quality: hygiene andenhancing factors”, Quality in Services II, selected papers, Warwick Business School,pp. 193-210.

Smith, W.L., Weatherly, K.A. and Tansik, D.A. (1992), “A customer-based service quality study:use of the sorting/cluster analysis methodology”, Proceedings of the 2nd InternationalResearch Seminar in Service Management, La Londes les Maures, June, pp. 591-603.

Stauss, B. and Hentschel, B. (1992), “Attribute-based versus incident-based measurement of servicequality: results of an empirical study in the German car service industry”, in Kunst, P. andLemmink, J. (Eds), Quality Management in Services, Van Gorcum, Assen, pp. 59-78.

Swan, J.E. and Combs, L.J. (1976), “Product performance and consumer satisfaction: a newconcept”, Journal of Marketing, Vol. 40, April, pp. 25-33.

Tan, K.C. and Pawitra, T.A. (2001), “Integrating SERVQUAL and Kano’s model into QFD forservice excellence development”, Managing Service Quality, Vol. 11 No. 6, pp. 418-30.

Tan, K.C. and Shen, X.X. (2000), “Integrating Kano’s model in the planning matrix of qualityfunction deployment”, Total Quality Management, Vol. 11 No. 8, pp. 1141-51.

Tan, K.C., Xie, M. and Shen, X.X. (1999), “Development of innovative products using Kano’smodel and quality function deployment”, International Journal of InnovationManagement, Vol. 3 No. 3, pp. 271-86.

Ting, S.C. and Chen, C.N. (2002), “The asymmetrical and nonlinear effects of store qualityattributes on customer satisfaction”, Total Quality Management, Vol. 13 No. 4, pp. 547-69.

Tontini, G. (2000), “Identification of customer attractive and must-be requirements using amodified Kano’s method: guidelines and case study”, Proceedings of the 54th AmericanQuality Congress, Indianapolis, IN, pp. 728-34.

Tontini, G. and Silveira, A. (2007), “Identification of satisfaction attributes using competitiveanalysis of the improvement gap”, International Journal of Operations & ProductionManagement, Vol. 27 No. 5, pp. 482-500.

Vavra, T.G. (1997), Improving Your Measurement of Customer Satisfaction: A Guide to Creating,Conducting, Analyzing and Reporting Customer Satisfaction Measurement Program,ASQC Quality Press, Milwaukee, WI.

Wiley, J.B., MacLachlan, D.L. and Moinpour, R. (1977), “Comparison of stated and inferredparameter values in additive models: an illustration of a paradigm”, Advances inConsumer Research, Vol. 4 No. 1, pp. 98-105.

Witell, L. and Lofgren, M. (2007), “Classification of quality attributes”, Managing Service Quality,Vol. 17 No. 1, pp. 54-73.

Yang, C.C. (2005), “The refined Kano’s model and its application”, Total Quality Managementand Business Excellence, Vol. 16 No. 10, pp. 1127-37.

About the authorsJosip Mikulic, PhD is a Teaching and Research Assistant in the Department of Tourism, Facultyof Economics and Business, University of Zagreb, Croatia. His primary research interests involvesatisfaction and loyalty research, particularly in the tourism and hospitality sector. He has

Classifyingquality attributes

65

published articles in Managing Service Quality and Tourism Management. Josip Mikulic is thecorresponding author and can be contacted at: josip.mikulic@efzg.hr

Darko Prebezac, PhD is a Professor of Tourism Marketing in the Faculty of Economics &Business, University of Zagreb. His research interests include marketing management, airlineindustry trends and analysis of tourism transport from a multidisciplinary perspective. He isHonorary Visiting Professor at the University of Ljubljana (Faculty of Economics) andUniversity of Hawaii (School of Travel Industry Management). He held high managerialpositions in the largest airline companies in the region for 12 years. He is a member of theUNWTO Education and Science Council and the International Association of Scientific Expertsin Tourism (AIEST).

MSQ21,1

66

To purchase reprints of this article please e-mail: reprints@emeraldinsight.comOr visit our web site for further details: www.emeraldinsight.com/reprints

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.