center for quality of management journalwalden-family.com/public/cqm-journal/2-4-whole-issue.pdf ·...

From the Chairman of the Editorial Board Page 2David Walden

A special issue onKano’s Methods for Understanding Customer-defined Quality

I. Introduction to Kano’s Methods Page 3

IIa. Developing and Administering Kano Questionnaires Page 7

IIb. Self-Stated Importance Questionnaire 2

IIIa. Experience in the Use of Kano’s Methods in theSpecification of BBN RS/1 Release 5.0 Page 12

IIIb. Observation of the Use of Kano’s Method Page 15

IIIc. A Bose Development Team’s Experiencewith Kano Mapping Page 15

IVa. The Desire for Continuous and Graphical Analysis Page 17

IVb. Thoughts on Graphical and Continuous Analyis Page 20

Va. Theoretical Parallels between Kano’sand Hertzberg’s Theories Page 23

Vb. Wording of the Kano Questionnaire Page 25

Vc. Theoretical Issues of Kano’s Methods Page 28

Volume 2, Number 4 Fall 1993© Copyright 1993, 1999 The Center for Quality of Management, Inc. Permission to make digital or hard copies of all or part of this work for personal or classroomuse is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the fullcitation on the first page. To copy otherwise, to republish, to post to servers, or to redistribute to lists requires prior specific permission and/or a fee.Copying is by permission of The Center for Quality of Management, Inc. • One Alewife Center, Suite 450 • Cambridge, Massachusetts 02140 USATelephone: (617) 873-8950 • Email: [email protected] The Center for Quality of Management Authors retain rights for re-publication of their articles.

ISSN: 1072-5296

CENTER FORQUALITY OF MANAGEMENT

JOURNAL

2 Fall 1993

From the Chairman of the Editorial Board

This issue is a compendium dedicated to the instruction, experience,ideas, and theories that have evolved from using Kano’s methods withinCQM companies. This paper has no single author. The material in the pa-per was pulled together from other CQM publications, from requests forexperiences of Kano users, and from theoretical ideas that have been circu-lating informally within the companies. I am grateful to all who contrib-uted to this paper and to those that allowed their writings to be adapted foruse in this paper.

The following individuals have refereed papers in 1993: David Boger,Dennis Buss, Ismael Dambolena, Alan Graham, Steve Graves, PaulHorwitz, Laurie Langone, Jeff Mayersohn, Joseph Posk, Ash Rao, DianeShen, and Robert Wood. I greatly appreciate their help. I also would like tothank those who helped with the production of the journal: Lene Abel,Barney Gage, Jay Howland, Trish McKinnon, Craig Murphy, MauryRingel, Ted Walls, and Robert Wood.

Fall 1993 3

Kano’s Methods for UnderstandingCustomer-defined QualityA compendium of ideas and experiences fromCharles Berger, Robert Blauth, David Boger,Christopher Bolster, Gary Burchill, WilliamDuMouchel, Fred Pouliot, Reinhart Richter,Allan Rubinoff, Diane Shen, Mike Timko, andDavid Walden

Useful ideas for improving the way we useKano’s methods are scattered throughout sec-tions II through V.

We hope that this compendium will helpreaders use Kano’s methods more successfully—and that the experience and wisdom gained in sodoing also can be shared through the CQM.

I. Introduction to Kano’s Methods2

In planning a product or service, one makes a listof potential customer needs that the product orservice should perhaps try to satisfy. Going tosee current and potential customers (“voice-of-the-customer” visits) is one good way to getideas for what should be on the list of potentialcustomer requirements. For simplicity, through-out the rest of this paper we will call these poten-tial customer requirements simply “customerrequirements” or “CRs,” with the implicit under-standing that one must investigate any CR list ingreater detail to understand which of the cus-tomer requirements need to be included in the fi-nal product (or service).

Many methods are available for investigat-ing the characteristics of customer require-ments.3 For instance, one can ask customers torank-order them. The particular method we willdiscuss here is based on the work of ProfessorNoriaki Kano of Tokyo Rika University.

Professor Kano and his colleagues developeda set of ideas4 that we summarize as follows.

1. Invisible ideas about quality can be made visible.

Customer ideas about quality are often confusedand difficult to see clearly, but they can be made

1 G. Burchill, D. Shen, et al., Concept Engineering Manual, CQMDocument 7I, September, 1992.2 Much of the text in this section is adapted, with permission, fromShoji Shiba et al., A New American TQM, Portland, Oregon, 1993,Productivity Press and the CQM, pp. 221-230.3 Chapters 7 to 11 of Design and Marketing of New Products byGlen L. Urban and John R. Hauser, 2nd edition, (Englewood Cliffs,NJ, Prentice Hall, 1993) describes many other methods of inquiringabout and understanding customer needs.4 Noriaki Kano et al., “Attractive Quality and Must-be Quality,”research summary of a presentation given at Nippon QC Gakka:12th Annual Meeting (1982), January 18, 1984. Presentations givenat Japanese Society for Quality Control Annual Meetings, NoriakiKano and Fumio Takahashi, “Himshitsu no M-H Sei Ni Tsuite”(Motivator and Hygiene Factor in Quality), October 1979; NoriakiKano, Shimichi Tsuji, Nobuhiko Seraku, and Fumio Takahashi,“Miryokuteki Himshitsu to Atarimae Himshitsu (1), (2)” (AttractiveQuality and Must-be Quality [1], [2]), October 1982, and publishedin Quality, JSQC, vol. 14, no. 2 (Tokyo: Japanese Society forQuality Control, 1984).

The Center for Quality Management (CQM) six-day introductory TQM course presents severalconcepts relating to understanding customer-de-fined quality based on the work of ProfessorNoriaki Kano of Tokyo Rika University and sev-eral of his colleagues in Japan. These conceptshave become embedded in the Concept Engi-neering (CE) process for operationally definingcustomer requirements,1 which a number ofCQM companies are using as part of their prod-uct development process.

When first introduced to Kano’s ideas,people are usually excited by them. Kano’sideas about types of quality are often a revela-tion, and his method for sorting the features of aproduct into various quality categories based ona questionnaire filled out by customers offers anapparently straightforward process for gainingdeep understanding of customer requirements.However, as with so many tools, successful ap-plication of Kano’s methods requires skill andexperience.

Within the CQM companies, we now have agood bit of experience with the application ofKano’s methods and some insight into how touse the methods effectively. This compendiumof material provides an overview of Kano’smethods, presents some experiences using themethods and some tips for use, and discusses anumber of subtleties that users should be awareof. It is divided into the following major catego-ries, most of which have more than one subsec-tion.

I. Introduction to Kano’s methods (page 3)II. Developing and administering Kano

questionnaires (page 7)III. Experiences in the use of Kano’s meth-

ods (page 12)IV. Continuous and graphical analysis of

Kano data (page 17)V. Theoretical issues relating to Kano’s

methods (page 23)

4 Fall 1993

clear. As the customer ideas of quality are madeclear, many requirements emerge, and they fallinto several groups, as represented by the treestructure of customer requirements in figure 1(below).

2. For some customer requirements, customersatisfaction is proportional to how fully functionalthe product is.

The horizontal axis of the Kano diagram in fig-ure 2 (above right) indicates how fully functionalsome aspect of a product is, and the vertical axisindicates how satisfied the customer is. Tradi-tional ideas about quality have sometimes as-sumed that customer satisfaction was simplyproportional to how functional the product was–that is, the less functional the product, the lesssatisfied the customer, and the more functionalthe product, the more satisfied the customer. Inthe figure, the line going through the origin at 45degrees graphs the situation in which customersatisfaction is simply proportional to how fullyfunctional the product is: the situation in whichthe customer is more satisfied (up) with a morefully functional product (right) and less satisfied(down) with a less functional product (left).

Such customer requirements Kano desig-nates as “One-dimensional” customer require-ments.5 For example, in automobiles, gas mile-age (unless it is quite bad) is likely to be a One-dimensional customer requirement: Better gasmileage provides more customer satisfaction andworse gas mileage provides more customer dis-satisfaction. (Study the graph for a moment andbe sure you understand how the 45-degree linethrough the origin indicates a One-dimensionalcustomer requirement, one where more function-ality leads to more satisfaction.)

3. Some customer requirements are not One-dimensional—there are also “must-be” and“attractive” elements.

Figure 2 also has curves labeled “Must-be” and“Attractive.”6 The Must-be curve indicates as-pects where the customer is more dissatisfiedwhen the product is less functional, but wherethe customer’s satisfaction never rises aboveneutral no matter how functional the product be-comes. For instance, having poor brakes in anautomobile causes a customer to be dissatisfied;having good brakes, however, does not raise thelevel of the customer’s satisfaction. Goodbrakes are expected—they are a Must-be re-quirement.7 (Study the Must-be curve to under-stand this requirement situation: A little or a lotof a given feature leaves the customer unmoved[basically neutral and certainly not satisfied];however, any lack of that particular featurequickly dissatisfies the customer.)

The Attractive curve indicates areas in whichthe customer is more satisfied when the productis more functional but is not dissatisfied whenthe product is less functional. For instance, anautomobile customer may not be unsatisfied ifthe radio antenna does not automatically loweritself into the car body when the radio is turnedoff, but the customer may be more satisfied whenthe car has this feature.8 (Study the Attractive

The Kano Diagram

One-dimensionalAttractive

Must-be

Customer Satisfied

Product Fully Functional

Product Dysfunctional

Customer Dissatisfied

Figure 2

5 Some companies use the word “satisfiers” instead of One-dimensional customer requirement. That is, the more fulfilled thisrequirement is, the more satisfied the customer is.6 These curves are just roughly sketched. Section Vc will considertheir exact shape in more detail.7 Some companies call these Must-be elements “dissatisfiers”; theycan dissatisfy but they cannot produce a significant level ofsatisfaction.8 Some companies call these attractive elements “delighters,” thatis, they do not dissatisfy if absent but can delight when present.

Example of Customer Requirements

Customer

requirements

for an

automobile

retractable radio antenna

rear window wiper

electronic door locker

gas mileage

warranty period

turning radius

brakes

windshield

engine cooling system

Figure 1

David Walden is asenior vicepresident of BBNand is generalmanager of BBN’sCommunicationsDivision. Hecompiled thematerial in thispaper —editorialcomments andunattributed textare by him.

Fall 1993 5

curve to understand the kind of situation wherelack of a feature leaves a customer basically neu-tral [certainly not dissatisfied]; however, havingmore of the feature quickly increases thecustomer’s level of satisfaction.)

A customer may also be indifferent to a qual-ity element, for instance to having a cigarettelighter in a car. Indifference would be plotted onfigure 2 roughly along the horizontal axis—thatis, the customer is neither satisfied nor dissatis-fied whether the product is dysfunctional or fullyfunctional.9

4. Customer requirements can be classified byquestionnaire.

Kano and his colleagues believe that the One-di-mensional, Attractive, Must-be, and Indifferentcustomer requirements can be classified througha customer questionnaire. On this questionnaireeach question has two parts: How do you feel ifthat feature is present in the product, and how doyou feel if that feature is not present in the prod-uct? To each part of the question, the customercan answer in one of five different ways.10 Forexample, see figure 3 (below).

Based on the responses to the two parts ofthe question in figure 3, the product feature (howgood gas mileage is, in this example) can beclassified into one of six categories:

A=AttractiveM=Must-beO=One-dimensionalI=IndifferentR=ReverselQ=Questionable

Figure 311

functional form ofthe question

If the gas mileage is good,how do you feel?

If the gas mileage is poor,how do you feel?

1. I like it that way.2. It must be that way.3. I am neutral.4. I can live with it that way.5. I dislike it that way.


dysfunctional formof the question

A Pair of Customer Requirement Questions in a Kano Questionnaire

The first four categories have already been de-fined, and these are primarily what we are seek-ing in the Kano analysis. The other twocategories indicate the following situations:There is a contradiction in the customer’s an-swers to the questions (=Questionable); or our apriori judgment of functional and dysfunctionalwas the reverse what the customer feels (=Re-verse).12

For each customer one determines intowhich category a given product feature falls bylooking up the customer’s answers to thatfeature’s questions in the following Kano Evalu-ation Table, figure 4 (next page).13

For example, if the customer answers, “1. Ilike it that way,” about “gas mileage is good,”the functional form of the question, and “5. I dis-like it that way,” about “gas mileage is poor,” thedysfunctional form of the question, we look atthe intersection of the first row and the fifth col-umn and find an O, indicating that gas mileage isa One-dimensional customer requirement from

9 The Kano diagram showing in figure 2 is useful for illustratingKano’s concepts of quality. However, the tabulations explained inpart 4 of this section are more useful for understanding the qualityelements of an actual product.10 These are the translation of Kano’s five answers as the CQMcompanies were taught them by Professor Shoji Shiba in 1990.Sections III and V discuss alternative formulations of the fiveanswers.11 The answers to the questions are numbered in this example to helpthe reader follow the description of how a Kano questionnaire isprocessed. Subsection IIIa gives a fairly compelling argument fornot numbering the answers.12 See part IIa.4 for a suggestion of how to handle the Reverse casein certain instances.13 A derivation of this table is explained in subsection Vc.

6 Fall 1993

Kano Evaluation Table

Figure 4

Examples of Three (Potential) Customer Requirements in a Kano Questionnaire

Figure 5


1a. If the gas mileage is good, how do you feel?


1b. If the gas mileage is poor, how do you feel?

2a. If the brakes are good, how do you feel?


2b. If the brakes are poor, how do you feel?


3a. If the radio antenna automatically retracts when the radio is turned off, how do you feel?


3b. If the radio antenna does not automatically retract when the radio is turned off, how do you feel?


Func-tional

1. like

2. must-be

4. live with

Q

R

R

A

I

I

A A

I I

II

O

M

3. neutral R I II M

M

5. dislike R R R R Q

Dysfunctional1.

like2.

must-be3.

neutral4.

live with5.

dislike

Customer Requirement is:A: AttractiveM: Must-beR: Reverse

O: One-dimensionalQ: Questionable resultI: Indifferent

CustomerRequirements

Fall 1993 7

the customer’s point of view.A portion of the Kano questionnaire for a

survey about an automobile is reproduced in fig-ure 5 (facing page).14

Once all the Kano questionnaires for a sur-vey have been collected, one tabulates them bylooking up the classification of each customerrequirement on each questionnaire in the KanoEvaluation Table and tallying it in the appropri-ate place in the row for that requirement on aKano questionnaire tabulation form (figure 6,next page).

The result of the tallying of all customer’squestionnaires is a matrix such as the one shownin figure 7 (next page).

For each row of the tabulation, that is, foreach customer requirement, the dominant cus-tomer view (dominate classification) is indicatedby the highest tally.15 If two or more categoriesare tied or close to tied, it may be an indicationthat more information is needed: You may bedealing with two market segments, or you mayneed to ask questions about more detailed cus-tomer requirements.

From the tabulation of customer responses tothe Kano survey for the automobile, a Kano dia-gram can be annotated–for example, a diagramshowing the Must-be, One-dimensional, and At-tractive, as in figure 8 (page 39)16.

All customer requirements are not equal.Improving performance on a Must-be customerrequirement that is already at a satisfactory levelis not productive when compared to improvingperformance on a One-dimensional or Attractivecustomer requirement. Having insight intowhich customer requirements fall into whichquality dimensions can improve focusing “on thevital few.”17

IIa. Developing and Administering KanoQuestionnaires18

In general, the steps to follow to develop and usea Kano questionnaire are:

1. Develop the questionnaire.2. Test the questionnaire and revise if necessary.3. Administer the questionnaire to customers.4. Process the results.5. Analyze the results.

1. Developing the Questionnaire

To construct the questionnaire formulate a pairof questions for each potential customer require-ment for which you desire customer feedback.In the Concept Engineering (CE) process, thesepotential customer requirements come from step6, the requirements KJ; the Kano process itself is

step 7 of Concept Engineering. In Concept Engi-neering, the Kano questionnaire and analysis areused to confirm and categorize the customer re-quirements discovered through voice-of-the-cus-tomer visits and analysis of the data collectedfrom customers. In other situations, potentialcustomer requirements might be developed inother ways. For instance, a survey might test as-pects of an existing product or service to find outif they are valid—if customers appreciate themas much as, or in the same way that, the providerof the product or service expects.

The first question in each pair of questionsfor a customer requirement refers to a situationin which the requirement is met, and it is wordedin a format similar to the following: “If [theproduct] satisfied [requirement x], how wouldyou feel?” This is the functional question. Thesecond question in each pair refers to the casewhere the requirement is not met. This is calledthe dysfunctional question, and is worded in aformat similar to the following: “If [the product]did not satisfy [requirement x], how would youfeel?”

When writing the pairs of functional anddysfunctional questions for each potential cus-tomer requirement, use the following guide-lines:19

• You may have to step down the ladder of ab-straction to construct a clear question.When the potential requirements comefrom analysis of voice-of-the-customerdata, avoid straying from the original intentof the customer requirement statement. InConcept Engineering, refer to the require-ment KJ and translation worksheets if nec-essary.

• Beware of polar wording in the questionpairs; multivalued orientation is preferred.Consider this functional question: “If lineplaced in the basket stayed in it until cast,how would you feel?” Instead of wordingthe dysfunctional question “If line placed inthe basket falls out before casting, how

14 In this example the first question in every pair is the functionalform of the question, and the second question in every pair is thedysfunctional form of the question.15 The mode of the distribution.16 This optional annotated Kano diagram may be useful inexplaining the results of the survey to other people; however, it hasa lot less information than a tabulation table such as that shown infigure 7.17 See part 5 of section II for more discussion on how to act on theKano results.18 Much of the text in this section is adapted with permission fromConcept Engineering, CQM Document 7I, G Burchill, D. Shen, etal., 1992, pp. 3.3-3.12.19 It is also helpful to get some advice from someone in your firmwho has experience in administering surveys.

Diane Shenworkedextensively todevelop the CQMConceptEngineeringmanual. She nowworks withvarious firms thatare implementingConceptEngineeringmethodologiesand is coursemanager for theCQM ConceptEngineeringCourse.

8 Fall 1993

Looking Up Questionnaire Answers in Evaluation Table and Tabulating the Result

Figure 6

Figure 7

Tabulation of Responses for Each Customer Requirement in a Kano Questionnaire about an Automobile

1.

2.

3.

21

23 M

23 O

13 5 5 23 A

6 1 4 1 11 23 I

1 9 6 1 6 23 M

7 2 3 1 10 23 I

C.R. A M O R Q I total grade

...

...

...

1 1

22 1

O

Dysfunctional1.

like2.

must-be3.

neutral4.

live with5.

dislike

Func-tional

1. like

2. must-be

3. neutral4. live with

5. dislike



1.

2.

3.

4.

1

C.R. A M O R Q I total grade

Tabulation of Surveys

1. I like it that way.2. It must be that way.3. I am neutral.4. I can live with it that way5. I dislike it that way.

1st CustomerRequirement—functional

form

1. I like it that way.2. It must be that way.3. I am neutral.4. I can live with it that way5. I dislike it that way.

Questionnaire

1st CustomerRequirement—dysfunc-tional form

Fall 1993 9

would you feel?,” it is preferable to ask, “Ifsome line placed in the basket falls out be-fore casting, how would you feel?”20

• Don’t try to cram several thoughts into onequestion. You want to know which ques-tion the respondent is answering. If a par-ticular requirement contains more than onethought, use more than one Kano question,but bear in mind the need to keep the surveyto a manageable length.

• Make sure questions are in customer terms,not development terms, that is, in terms ofbenefits, not features.

• The time you are taking to hear your cus-tomers’ views contributes to the company’sprofessional image. The format of thequestionnaire should further enhance thatimage.

2. Testing the Questionnaire

When a questionnaire is to be sent to many cus-tomers, it’s important that it be understandable.This is especially true of a Kano questionnaire,since it is unfamiliar to most people asked to fillit out. We recommend testing all of your ques-tionnaires internally before distributing them tocustomers. A test run will help identify unclearwording, typographical errors, or confusing in-structions.

Refining the questionnaire may require itera-tion. These guidelines can help you test your

questionnaire effectively:• Have members of the team preparing the

Kano survey answer the questionnaire first.Each team member should think of a cus-tomer, try to predict his or her response,and note which questions the customer maynot understand.

• Next select people inside your company toanswer the questionnaire, and administer itto them. Select from a variety of back-grounds, for example, senior managers, de-velopment engineers, marketing personnel,and so on.

• Revise the questions and retest.• Listen carefully and without bias to what

your internal test customers say. If theyfind something confusing, it is quite likelythat others will as well. Revise questions oradd instructions as necessary.

3. Administering the questionnaire

The points below suggest some of the many de-tails to consider in administering your question-naire. You may be able to prepare for several ofthe points long before you have to prepare theKano survey, thus shortening the time to preparethe questionnaire, send it, and wait for re-sponses.

Kano Diagram for Car Customer Requirements

Figure 8

20 This example is from the case study of the stripping basket forfishing, described in the Concept Engineering Manual.

Attractive:retractingradioantenna

Must-Be:good brakes

1-Dimensional: gasmileage

Dysfunctional

Satisfaction

Dissatisfaction

Functional

10 Fall 1993

• Select the customers you would like to sur-vey. If you are using Concept Engineering,we suggest returning to the customer selec-tion matrix to develop a target list, applyingthe same criteria used there. In order to en-sure a statistically significant sample, mostteams poll considerably more customersthan were interviewed. Remember that notall customers will respond.

• Decide what medium you want to use: tele-phone (voice or fax), face-to-face presenta-tion, mail, or other means. The mostcommon method is through the mail. If youplan to use the mail, write a letter of intro-duction that explains the purpose of the sur-vey and includes directions for thecustomer.

• Collect demographic data that will enableyou to distinguish potential market seg-ments if they exist. Categories of helpfulinformation might include company andpersonal characteristics, familiarity or ex-perience with product, use of competitor’sproducts, and so forth.

• Include instructions for filling out the ques-tionnaire. This is particularly important forthe Kano questionnaire, since it is likely tobe new to customers. Subsection Vb pro-vides sample instructions.

• If you are using a Self-stated Importancequestionnaire21 in addition to the Kanoquestionnaire, use the same sequence ofquestions in both to make comparing the re-sults of the two questionnaires easier.

• Send out the survey. Keep a log of custom-ers to whom the survey was sent, alongwith the date. This will help you follow up,if necessary; and will help you avoid dupli-cation if you choose to expand your samplelater.

• Record responses as they arrive.

4. Processing the results

In part 4 of section I, we sketched how to processeach completed questionnaire by looking up, inthe Kano Evaluation Table, a “score” for the an-swers to the pair of questions for each potentialcustomer requirement being investigated. Weexplained how these scores are in turn talliedinto a tabulation matrix. Figure 6 illustrates thisprocess.

In the far right column of the tabulation ma-trix, assign a category (grade) to each of the re-quirements. The simplest way to choose acategory is to use whatever code appears mostoften in the responses for that requirement (that

is, to use the statistical mode of the responses).22

If any requirement receives a substantialnumber of Questionable (Q) scores, it shouldprobably be temporarily deleted from the analy-sis until the confusion with the question can beresolved or the thought processes of the respon-dents can be explored.

If a majority of the people who responded tothe questionnaire give a Reverse (R) score to oneof the customer requirements, it indicates thatthe marketplace’s thoughts about that questionare the opposite of those of the survey’s creators.For instance, a company providing packaged va-cation tours might construct a pair of questionssuch as the following:

Functional: How do you feel about a vaca-tion package where there are pre-plannedevents most of every day?Dysfunctional: How do you feel about a va-cation package where there are few pre-planned events each day?

If only a few people respond that they dislikehaving preplanned events most of every day andlike having few preplanned events each day (seecell 5-1 in figure 4), this indicates that there are afew people in the marketplace who are differentfrom most other people. (There may be a goodmarket segmentation opportunity here.) But, if alarge number of the responses say they dislikehaving preplanned events most of every day andlike having few preplanned events each day, thislarge number of reverse scores indicates that thequestionnaire’s creators are thinking aboutthings in the reverse of the way most customersare thinking about them.

In this case, one can rescore all of the ques-tionnaires for this pair of questions, using the an-swers to the dysfunctional question as an indexto the functional side of the Kano EvaluationTable, and the answers to the functional ques-tions as an index to the dysfunctional side of theKano Evaluation Table. This retroactively con-verts the question into the reverse form favoredby most customers:

Dysfunctional: How do you feel about a va-cation package where there are preplannedevents most of every day?Functional: How do you feel about a vaca-tion package where there are few preplannedevents each day?

In this way, the data already collected can beused. To repeat, to make this retroactive conver-

21 See description in subsection IIb.22 See section IV for more sophisticated analysis methods.

Fall 1993 11

sion work, the responses for all people who an-swered this pair of questions must be rescored,whether or not their original score for the pair ofquestions was Reverse.5. Analyzing the Results

Several benefits are obtained from analyzingKano data:

• Gaining a better understanding of require-ments;

• Prioritizing requirements for developmentactivities;

• Distinguishing market segment characteris-tics;

• Aiding in the design tradeoff process.The purpose of administering a Kano ques-

tionnaire is to understand better the characteris-tics of customer’s requirements. The responsesshould be seen only as a guide; they do not pro-vide exact answers as to which features must beincluded in the product or which requirementsneed not be fully satisfied.

Establishing the method by which you willanalyze the data before disseminating the ques-tionnaires will save time. Will you use manualor automated input and analysis tools? Knowingthis will enable you to marshal the necessary re-sources while you are waiting for replies.

There are various possible analysis ap-proaches. As mentioned earlier, a simple way torank the requirements is to score each accordingto the mode (the most frequently occurring di-mension) in each row of the tabulation matrix.Thus, a requirement voted Must-be (M) by 43percent of respondents and Attractive (A) by 38percent of respondents would be interpreted as aMust-be requirement.

You may want to go beyond the simple modeand look at the second most frequent dimensionfor each requirement. For example, take a casewhere there are two questions and fifty re-sponses to each. Thirty customers rate the firstrequirement Attractive and twenty rate it Indif-ferent. On the second requirement, thirty cus-tomers again rate it Attractive, but the remainingtwenty rate it Must-Be. In this case, it is likelythat the two requirements should be treated dif-ferently. The second requirement should receivehigher priority from the team.

When two Kano codes are tied in the scoringfor a given question, consider:

a. Following up with customers for additionalinsight,

b. Looking for market segmentation differ-ences,

c. Selecting the classification that would have

the greatest impact on the product (use thefollowing ordering: M > O > A > I).

Another way to study the data is to constructa spreadsheet with columns for the first, second,and third most frequent responses, as in figure 9(below).

Spreadsheet of Most Frequent Responses toCustomer Requirements

Then the rows can be rearranged into groupsaccording to the following order: Must-Be’sfirst, One-Dimensionals second, followed byAttractives, Indifferents, and Reverse require-ments, as in figure 10 (below).

Spreadsheet of Customer RequirementsSorted in Order of Most Frequent Response

Mostfrequentresponse

A

Customerrequirement

number

1

Second mostfrequentresponse

A2

I

3

8

5

6

7

9

4

M

M

O

M

A

O

O

A

A

O

M

I

M

I

Third mostfrequentresponse

O

O

Figure 9

Mostfrequentresponse

M

Customerrequirement

number

8

Second mostfrequentresponse

M3

I

6

7

5

4

1

2

9

M

A

O

O

A

A

O

A

A

O

O

M

O

A

M

I

Third mostfrequentresponse

Figure 10

12 Fall 1993

If you used a Self-stated Importance Ques-tionnaire23 in parallel with the Kano survey, itcan be helpful in further sorting the Kano re-sponses. One way to order the requirementswithin each group is by importance ranking. Forexample, if there were several customer require-ments whose mode was Attractive, you mightuse the Self-stated Importance data to rank theAttractive requirements in descending order ofimportance. This would help you discern whichAttractive requirements the development teammight want to focus on.

The Kano survey results will be only one ofseveral factors that will dictate what should beincluded in a product. A general guideline is toseek to fulfill all Must-be requirements, be com-petitive with market leaders on the One-dimen-sional Requirements, and include somedifferentiating Attractive elements.

In general, the return you can expect fromfulfilling a requirement (in terms of customersatisfaction) should guide the effort you invest tofulfill it. As stated earlier, improving perfor-mance on a Must-be customer requirement thatis already at a satisfactory level is not as produc-tive as improving performance on a One-Dimen-sional or Attractive requirement. Classifyingcustomer requirements into Kano’s dimensionswill allow you to focus your efforts where yourcustomer will notice their effect the most.

IIb. Self-Stated ImportanceQuestionnaire24

The CQM companies using Kano’s methods of-ten use a Self-stated Importance questionnairetogether with the Kano questionnaire. There-fore, we describe it here.

According to research by Professor John R.

Hauser at MIT,25 the Self-stated Importancequestionnaire can help organizations understandthe relative importance of each requirement forcustomers. An organization can use such a sur-vey in parallel with a Kano survey to help focusattention on the most important results from theKano survey.

Constructing the Self-Stated Importancequestionnaire is straightforward:

1. For each of the potential customer require-ments to be included in the Kano question-naire, construct a question on the Self-statedImportance questionnaire in the followinggeneral format: “How important is it orwould it be if: [requirement x]?” For ex-ample: “How important is it or would it be ifthe car has good gas mileage?”2. Provide a scale on which customers canmark their responses from “Not at all impor-tant” to “Extremely important.” Figure 11(below) gives an example.

IIIa. Experience in the Use of Kano’sMethods in the Specification ofBBN RS/1 Release 5.0[Editor’s Note: Robert Blauth, Reinhart Richterand Allan Rubinoff of BBN Software Products

Figure 11

Example from Self-stated Importance Questionnaire

23 See subsection IIb, on this page.24 The text in this section is adapted from G. Burchill, D. Shen, et.al. , Concept Engineering. CQM Document 71, 1992, pp. 3.3-3.4.25 John R. Hauser and Don Clausing, “The House of Quality,”Harvard Business Review No. 3 (May-June 1988); John R. Hauser,“Comparison of Importance Measurement Methodologies and theirRelationship to Consumer Satisfaction,” MIT Marketing CenterWorking Paper 91-1, 1991. Chapters 7 to 11 of “Design andMarketing of New Products” by Glen L. Urban and John R. Hauser(Englewood Cliffs, NJ, Prentice Hall, 1993) describe many othermethods of inquiring about and understanding customer needs. Thatbook uses the term Self-rated Importance questionnaire.

How important is it or would it be if:The car has good gas mileage? 1 2 3 4 5 6 7 8 9

How important is it or would it be if:The car has good brakes? 1 2 3 4 5 6 7 8 9

How important is it or would it be if:The car has a long warranty period? 1 2 3 4 5 6 7 8 9

How important is it or would it be if:The car has a small turning radius? 1 2 3 4 5 6 7 8 9

Important Not at all important

Somewhatimportant

Veryimportant

Extremelyimportant

Fall 1993 13

were asked to provide some informal thoughtson the use of Kano’s methods in connection withspecifying Release 5.0 of BBN’s 10-year old RS/1 statistical data analysis product.]

Reinhart Richter says: As I understandKano’s method, it is a method of using five stan-dard answers (“Like,” “Must-be,” etc.) to func-tional and dysfunctional questions about aproduct feature or function to classify (not rank)the feature or function into one of six categories:

“Must be” meaning if the product does nothave this no one will be interested in it.“One-dimensional” meaning the more youprovide this function the more satisfied thecustomer will be“Attractive” meaning the customer is happywhen it is there but will not complain if it isnot there“Indifferent” meaning the customer does notcare about this feature“Questionable” and “Reverse” meaning thatwe probably have done something wrongphrasing the question or, under certain cir-cumstances, that this particular function hasa negative interaction with other importantfunctions and we have not really understoodwhat we are talking about.

I personally think that this method is bril-liant. It does have some intrinsic difficulties,though.A. The person answering the questions needs tounderstand that the default answers shall reflecta classification, not a ranking. If the answersare misunderstood as being a ranking on a 1 to 5scale, they may misunderstand how to interpretthe answers, and then analysis with the 5x5 KanoEvaluation Table may give misleading results.In particular, avoid numbering the default an-swers on the Kano questionnaire, and provideexamples of how to answer.B. Phrasing the standard answers to the questionshould be done with great care, particularly inan international environment. If you do nottranslate the questions and answers, people witha less profound knowledge of the English lan-guage may not understand them, and the resultswill be meaningless.

The first time our international customerbase was confronted with a Kano survey, theythought “It must be that way” was much strongerthan “I like it that way.” They were confusedabout why “It must be that way” was the second,not the first default answer. I myself understood

what this was all about only after studying up onit, and I was a member of the team administeringthe survey. Educate the people answering thequestions.26

If you do translate the questions and an-swers, on the other hand, you need to make surethat Kano’s “spirit” — that is, classification —survives.C. The mode statistic used to analyze the resultsmay need to be modified. If the Kano question-naire asks about very general functions, such aswhether a car should have three or four wheels orwhether a software package should have agraphical user interface or a command interface,everybody will have a specific opinion. If, how-ever, the Kano survey asks about very specificfunctions, such as whether a car should have ce-ramic valves or a software package should sup-port a particular printing device, then themajority of the respondents may be Indifferent.

Extremely detailed questions can increasethe “noise level” to a point where all “require-ments” are considered Indifferent. For example,if 18 answers classify a function as One-dimen-sional, 19 as Attractive, 18 as Must-be, 20 as In-different, 2 as Reverse, and 3 as Questionable,then the mode statistic classifies this function asIndifferent even though 57 out of 82 people an-swering say that they need this function in oneway or the other. Consider doing something todecrease the noise level before applying themode statistics.

One way to modify the mode statistic is asfollows:

In figure 12 (next page), for CR1,(19+18+18) > (20+2+3). Therefore, the grade isA; that is, maximum(18, 19, 18). For CR2,(7+6+9) < (36+2+2); therefore, the grade is I, ormaximum(36, 2, 2).

D. It may happen that the person answering theKano survey is already using a competitor’sproduct in addition to your product for certaintasks presently not covered by your product. Inthis case the person may answer the questionpairs with “Must be/Live with” for a particularfeature. The customer thinks that the function isvery important, but if it is not covered by your

If (One-dimensional+Attractive+Must-be) > (Indifferent+Reverse+Questionable),Then grade is maximum(One-dimensional, Attractive, Must-be),Else grade is maximum(Indifferent, Reverse, Questionable).

26 Subsection Vb includes an example of instructions on how toanswer questions.

Robert Blauth,Reinhart Richterand AllanRubinoff areemployees ofBBN’s SoftwareProductsDivision, andhave used Kano’smethods in thespecification ofproducts.

14 Fall 1993

with support for a 24 hour clock format and thefunctional version of the requireent was to pro-vide support only for a 12-hour format.

In formulating the questions, we did try toemphasize benefits rather than specific features,as the CQM instructions on Kano suggested.However, we were not always able to do so, andas a result a lot of the questions were fairly spe-cific about features. Some of this may have beenthe result our team’s unfamiliarity with Kanosurveys. But it may be that in some ways theKano survey was unsuitable for our purposes.Some of the things we were proposing requiredtalking about specific features; just emphasizingbenefits would have resulted in questions thatwere too vague to elicit useful feedback.

There should be a brief explanation of theKano format included with the questionnaire.This would help the customers better understandthe use of the positive and negative questions aswell as the meaning of the five possible answers.

The survey seems to be well suited to defin-ing a new product, but awkward for developing anew release of an existing mature product likeRS/1. For example, if a current customer says aspecific proposed feature is a Must-be, how dowe interpret that? That is, if the feature is trulyessential, why is that person using the product inthe first place, since the product doesn’t have thefeature now?

The survey was not very useful for distin-guishing between small features and large ones.For example, if a small feature is considered aMust-be by respondents, that’s probably notequivalent to a large feature being considered aMust-be. Presumably, a small feature won’tmake or break a sale of the product, whereas alarge feature might.

The survey helps you make decisions aboutwhat features to include in a product but is notvery helpful for determining how to implement afeature. Different people will have differentpreferences about how a feature should work,but it is hard to use the Kano survey to determinethis.

It seems that a Kano survey would be mostuseful for defining a new product, since it helpsyou determine what things a product should dowithout committing you to a specific approach todoing those things. In the case of RS/1 5.0,though, there were a lot of historical constraintson what the new release could do and how itcould do them.

Example of the use of modified mode statistic.

1.

2. I

A

C. R. A total grade

...

19 318 18

O M I R Q

90

7 6 9 36 2 2 62

20 2

Figure 12

27 See also the suggestions in subsection Vc for other changes in theKano Evaluation Table.

product it does not cause big problems, becausethe customer can use the competitor’s product.In this case the Kano Evaluation Table yields anIndifferent classification even though it shouldbe a Must-be. Consider modifying the KanoEvaluation Table element (functional=Mustbe,dysfunctonal=Live with) from Indifferent toMust-be.27

E. The Kano method should always be combinedwith a Self-stated Importance ranking.F. For existing products such as RS/1, use ofKano’s method to analyze the need of new func-tionality may be questionable. If a certain addi-tional function does not and cannot interferewith any other function or property of the soft-ware, then this additional function should not beable to be “Reverse” or “Questionable” unlessyou misphrased the question. Also, such an ad-ditional function should not cause a feeling of“Live with” or “Dislike.” Drop default answerson questions where they don’t make any sense.They probably only confuse the reader.

[Robert Blauth and Allan Rubinoff confirmsome of Richter’s points and make some addi-tional points.]

The wording of the five possible answers is themost problematic part of the survey. We ran intoproblems with international customers’ interpre-tation of “I like it that way” vs. “It must be thatway.” We also had trouble with the answer “Itmust be that way” to the dysfunctional version ofthe requirement statement—it is difficult forsome people to comprehend how “It must be thatway” can be a possible answer to the statementthat the product doesn’t have what the customerwants.

Wording the functional and dysfunctional re-quirement statements was difficult. Some of thefeatures that we asked about were not clear topeople answering the survey, and so the validityof some of the answers was questionable. Also,wording was difficult if the choice was really notblack or white. For example, the dysfunctionalversion of the requirement was to provide users

Fall 1993 15

It was important to look at combinations ofscores before making a final determination ofpriorities. For example, if one requirement fellinto the low end of the Attractive scores but hada large number of One-dimensional scores, wemight consider keeping that requirement in theAttractive list.

IIIb. Observation of the Use of Kano’sMethod

[Editor’s Note: As part of his PhD research,which included Concept Engineering, GaryBurchill has observed Kano’s methods in use asmuch as or more than anyone else associatedwith the CQM. He has provided the followingobservations.]I am absolutely convinced that characterizing thecustomer requirements into Kano’s categories isvery valuable. Design is a trade-off activity.Any credible insight the designer can developwith respect to which customer requirements canor cannot be sacrificed is invaluable for keepingdevelopment effort focused, making necessarytrade-offs, and maximizing customer benefit.

Here are some situations I have observed:

• Developers insisting that a particular re-quirement must be satisfied in the product,since it was a Must-be

• Program managers deciding on the basis ofrelative priorities within the Attractive cat-egory which requirements not to address inthe current release of the product

• A development team negotiating with itsmanagement about development schedulein order to accomodate key Attractive re-quirements

• Kano analysis (with relative priorities)seeming to increase the credibility and con-fidence of the team with respect to the de-sign objectives, perhaps on account ofincreased traceability

Writing the customer requirements in thefunctional and dysfunctional format of the Kanoquestionnaire is hard, however. The writer of apair of questions is forced down the ladder of ab-straction. This can create a situation where thepair of questions is about a smaller customer re-quirement than was heard in the voice-of-the-customer interviews. Writing more pairs ofquestions to address the remaining aspects of therequirement heard from the customers is not fea-sible, because it would result in too long a ques-tionnaire.

Writing multivalued pairs of functional and

dysfunctional questions is also difficult. To writea 0/1 pair of questions is easy, but when you de-velop question pairs that are anchored on oppo-site ends of the scale, the responses you receiveare often questionable.

Some people get distracted by the wordingand order of the five standard answers. This isunfortunate, since one individual on the teampreparing a Kano questionnaire with a prove-it-to-me attitude can disrupt progress of the entireteam. In the larger scheme of things, I am notsure it matters that much what words you use todescribe the five responses, especially when re-spondents presist in perceiving them to be an or-dinal scale from better to worse.

Adding the Self-stated Importance informa-tion increases the designer’s ability to discrimi-nate among customer requirements. MikeTimko’s extension to combine the two scalesinto “better-than” and “worse-than” scales is asiginficant enhancement.28

Engineers, in general, seem to have lesstrouble than marketing people in seeing the in-herent usefulness of Kano’s characterization ofthe various customer requirements and in over-looking the potential problems with the responsescale. Possibly this is because marketing peoplehave other methods for analyzing customer re-quirements with which they are already fullyconversant.

IIIc. A Bose Development Team’sExperience With Kano Mapping[Editor’s Note: Editorial Board member DavidBoger also had observations on the use of Kano’smethods.]At Bose we have used Kano’s methods as part ofthe use of Concept Engineering to plan a newproduct. Once the development team has soughtand analyzed the voice of the customer and de-veloped a list of key customer requirements, theteam must operationalize what it has learned.This means (a) defining metrics against whichsolutions for each customer requirement can betested, and (b) verifying that the team has accu-rately heard and understood the voice of the cus-tomer and translated it into an appropriaterequirement statement.

It is this latter element of Operationalizationthat employs the Kano’s techniques. Kano letsus categorize each requirement according to howcustomers define the relationship between thedegree to which the requirement’s functionality

Gary Burchill wasthe leadingcontributor in thedevelopment ofCE. Gary is nowcontinuing servicein the U.S. Navy,and is based inPennsylvania.

28 See subsection IIIc for a further endorsement of Timko’s ideasand subsection IVb for a description of this extension.

David Boger iswith BoseCorporation andparticipated inthe Enchiladacase studypresented in theSummer 1993issue of the CQMJournal.

16 Fall 1993

is present in the product concept and thecustomer’s perceived level of satisfaction. TheKano survey data gives the development team abasis for decisions with regard to leveraging re-sources. Essentially the team translates theKano results into a customer requirementweighting factor.

However, how best to make this translation?Is it reasonable simply to categorize each of therequirements into one of the four Kano dimen-sions? The Center for Quality Management’sConcept Engineering manual describes onemethod of taking into account the proportion ofscores in each dimension. Page 5.11 of the docu-ment shows how a team should redefine Kanoresults in order properly to complete a Pugh-style selection matrix.

An alternative approach for deriving aweighting factor has been suggested by a devel-opment team from Analog Devices. This ap-proach interpolates within the Kano Diagram’stwo dimensions of functionality and satisfaction,thereby providing more accurate weighting in-formation.29 The method actually arrives at twoindependent factors—one to be applied when thesolution under evaluation provides functionalitywith regard to a customer requirement betterthan a reference solution, and the other to be ap-plied when the solution under evaluation isworse than a reference solution.

Bose development teams have plotted theseso-called “better than” versus “worse than”scores and found the graphical representation il-luminating. Consider the graph in figure 13 (be-low).30

Two dimensional Representation of the Integrationof the Kano Quality Categories

In this example, five Customer Require-ments are plotted on the “better than” and “worsethan” axes. Points that lie near the origin (CR5)show low scores on each scale and are thereforeincreasingly Indifferent as they approach the ori-gin; these are the least important requirementsfor which function should be provided in theproduct concept. Points that lie farther from theorigin and hug the horizontal axis (CR1) are in-creasingly more pure Must-be Kano require-ments and therefore deserve the full attention ofthe development team. Points farther from theorigin and close to the unity slope line are in-creasingly one-dimensional (CR 3, 2). Finally,those points that reside away from the origin andclose to the vertical axis are more like Attractiveelements (CR4).

In general, it can be seen that the priority ofthe Customer Requirements may be representedin a descending order on this graph by a curvedline beginning on the “worse than” axis far fromthe origin and then sweeping in an arc in a coun-terclockwise direction ending on the “betterthan” axis near the origin. This is illustrated infigure 14 (below).

Order of Decreasing Importance

[Editor’s Note: I suspect there are cases when anAttractive requirement might be included in theproduct even when insufficient resources areavailable to meet adequately all the One-dimen-sional requirements in a first release. Maybecustomers would be more accepting of some

29 It may be useful to read Section IVb first for a derivation of thisapproach.30 Ibid.Figure 13

One-

dimensionalAttractive

IndifferentMust-be

Ifbetterthan

If worse than

CR1

CR2

CR3

CR4

CR5

Figure 14

One-

dimensionalAttractive

IndifferentMust-be

Ifbetterthan

If worse than

CR1

CR2

CR3

CR4

CR5

Fall 1993 17

weaknesses in meeting One-dimensional re-quirements if/when an occasional “delighter”were provided.]

IVa. The Desire for Continuous andGraphical AnalysisMany people who have used Kano’s methods, orare contemplating doing so, are wary of thesimple method given in sections I and II for clas-sifying a given customer requirement as Attrac-tive, One-dimensional, and so on. Theytypically feel that it seems like a loss of informa-tion (and in some cases is misleading) to reduce25 possible combinations of answers from eachrespondant into just one of six possible catego-ries, and then to reduce further the category de-rived for each respondent to just one of sixpossible categories for all respondents.

Thus, there has been a desire for more “con-tinuous” (finer) methods of analyzing the datafrom Kano questionnaires. In some cases themore continuous methods have been subjectiveand qualitative; for example, using commonsense to choose the Kano category for a particu-lar customer requirement when the category se-lected by the mode seems misleading. In othercases people have suggested and used morequantitative methods. The next two subsections,IVb and IVc, describe several of these.

After a quantitative, more continuousmethod for analyzing Kano questionnaires is de-rived, a natural next step is to show the resultsgraphically, since graphical methods are oftenbetter to indicate trends (which may be hidden ina forest of individual numbers) and to show moredimensions of the data simultaneously than canoften be comprehended from the raw numbers.

Thus, using what we are calling continuousvariables and representing the results graphicallyoffer at least three powerful advantages:

• A continuous approach can summarize thedata without losing resolution. For ex-ample, in the Kano Evaluation Table thereare nine response pairs that equate to theIndifferent category, and each may have asomewhat different emphasis.

• A continuous representation deals morecomfortably with situations where there isno dominant response to a question (e.g.,37 percent Must-be, 33 percent Attractive,30 percent One dimensional) by allowingfor intermediate points, or hybrids. See thecaveat below, however.

• A graphical representation can often con-vey more information simultaneously thana nongraphical approach.

The more continuous approaches describedin subsection IVb may be best applied after youinspect responses for evidence of discrete marketsegments. Lumping together distinctly differentperspectives to form an average may produceconfusing results. Take, for example, a casewhere votes are evenly split between Attractive,Must-be, and One-dimensional. One way tocheck for the existence of distinct segments is torun a test of correlation between this responsevariable and some of the demographic data col-lected with the questionnaire. When differentsegments are identified, it may be best to handlethe data separately. (Of course, the same consid-eration could be given to processing of the Kanodata using the simplest of methods.)

It also would be useful for someone with astrong mathematics and extensive testing back-ground to consider and explain the relationship,if any, between Kano-type questionnaires andmore mathematical techniques of anaylzing andorganizing customer analysis, such as, clusteranalysis, multidimensional scaling, or corre-spondence analysis.

IVb. An Experiment in ContinuousAnalysis[Editor’s Note: Mike Timko of Analog Deviceshas generously provided a blow-by-blow de-scription of some ideas he and his colleagues atAnalog Devices used to try to improve process-ing of Kano survey responses. At first he washesitant to provide a description of what he con-siders to be a trial use of some ideas. However,he was able to be convinced of the value of mak-ing this informally available so that others cantry to build on his work or find ways to improveit.]

During the past year I have been involved ina voice-of-the-customer project in the ConverterOperating Group (COG) at Analog Devices, ex-ploring “12-bit general purpose ADC” productdefinitions. There were about six of us on theteam, with Gary Burchill as our facilitator.

When the Kano questionnaires came back, Ivolunteered to put the results in a form that theteam could use. During that process, I came upwith a way to crunch the data that is easier toimplement while, I think, preserving the voice ofthe customer a little better than simply scoringthe answers as Attractive, Must-be, and so forth.

First, the Kano questionnaires were typedinto a computer program, written in C by ReddiPenumalli of our CAD department. This pro-gram took the two answers to each pair of ques-tions and looked up the resulting score

Mike Timko is adivision fellow atthe Semi-conductorDivision ofAnalog Devices.His respon-sibilities includethe definition,design anddevelopment ofhigh resolutionAnalog-to DigitalConvertors forthe dataacquisitionmarket. He holdsSB and SMdegrees in EEfrom MIT.

18 Fall 1993

31 A representation of Better and Worse can also be overlaid on thetraditional Kano diagram (see figure 2; however, we will not hereillustrate and describe the derivation of this overlay.

(Attractive, Must-be, etc.) in the Kano Evalua-tion Table. Another program then compiled theresults so that the data I got was a histogram ofthe number of responses in each Kano dimensionfor each question, plus an average importancevalue for each question from the Self-stated Im-portance questionnaires that were gathered at thesame time as the Kano questionnaires.

I wanted to put this information into aspreadsheet where it could be manipulated invarious ways, as our development team saw fit.My goal was to be able to compile a score for thevarious solutions the team had come up with bywriting spreadsheet functions (in Excel on aMacintosh).

I immediately encountered a problem. Thebasic Kano guidelines in the Concept Engineer-ing manual assume some use of judgment inclassifying each customer requirement into oneof the Kano dimensions by the majority of the re-sponses. The examples shown in the book are allpretty clear-cut. However, many of our re-sponses tended to be spread out over several cat-egories (A, O, M, and I)—see CR1 in the figure15. Simply using the mode statistic didn’t seemappropriate. Also, it didn’t seem quite fair toclassify a customer requirement that was 90 to10 Attractive to Indifferent the same as one thatwas 60 to 40 Attractive to Indifferent, as illus-trated by CR2 and CR3 in figure 15 (below).

Examples where Mode Statistic seems Inadequate

I wanted to calculate an average of some sortwhile also preserving some idea of the spreadover Attractive, One-dimensional, and Must-befeatures. This gave me the idea to reduce thedata to two numbers: a positive number that isthe relative value of meeting this customer re-quirement (versus the competition), and a nega-tive number that is the relative cost of notmeeting this customer requirement. I labeledthese columns “If we’re better ...” and “If we’reworse ...”, abbreviated Better and Worse.

To calculate the Better value, I added the At-tractive and One-dimensional columns and di-vided by the total number of Attractive,

One-dimensional, Must-be, and Indifferent re-sponses. (That is, I ignored Reverse and Ques-tionable responses.) I then added the Must-beand One-dimensional columns, divided by thesame normalizing factor, and put a minus sign infront of the result to get the number for Worse.These may be written as the following formulas:

Example where Better and Worse have beencalculated for Seven Customer Requirements

The positive Better numbers are indicative ofthe situation where, on average, customer satis-faction will be increased by providing these (At-tractive and One-dimensional) elements. Thenegative Worse numbers are indicative of thesituation where customer satisfaction will be de-creased if these (One-dimensional and Must-be)elements are not included. To look at it from aslightly different angle, Better indicates howmuch customer satisfaction is increased by ourproviding a feature (i.e., “If we’re better” [thanthe competition], increase our score by this posi-tive number); and Worse indicates how muchcustomer satisfaction is decreased by our notproviding the feature (i.e., “If we’re worse”[than the competition], decrease our score bythis number).

Pairs of Better and Worse points for eachcustomer requirement can be plotted on a two-di-mensional graph as shown in figure 17 on thenext page (the minus sign in front of Worse hasbeen ignored in this graph for purposes of clar-ity).31

100 A

100 A

C R A M O R Q I total grade

90

3.

33

10

1621 30

60 40 100 A

1.

2.

Figure 15

CR A M O I Better Worse

6

5

4

3

2

1

7

53

40

59

55

69

40

63

20

17

8

3

4

2

1

35

33

26

20

16

2

1

6

24

18

30

27

60

20

.77

.64

.77

.69

.73

.40

.75

-.48

-.44

-.31

-.21

-.17

-.04

-.02

Figure 16

Better= A + O

A + O + M + I Worse= − O + M

A + O + M + I

Fall 1993 19

Two Dimensional Representation of Kano QualityCategories

In the graph in figure 17, Better runs from0.0 to 1.0 up the vertical axis, and Worse runsfrom 0.0 to 1.0 along the horizontal axis (remem-ber, the minus sign has been left off Worse onthis graph). If there are 20 responses regarding a

particular customer requirement, and all of themwhen looked up in the Kano Evaluation Table re-sult in an Attractive rating, then Worse=(0+0)/(20+0+0+0)=0 and Better=(20+0)/(20+0+0+0)=1, or the point 0,1; this point is plotted at thetop left corner of the graph, the purely Attractivecorner.32 This and a number of other points aredescribed in figure 18 (below).

Finally, I decided to multiply each of thesenumbers by the Self-stated Importance average.If the Self-stated Importance scale runs from 0 to1.0,33 then potential requirements judged less im-portant will be scaled toward the Indifferent cor-ner of the graph shown in figure 17.

The team then used the scaled Better andWorse numbers in two ways.

First, most of the features on our customerrequirements list are independent of one another.That is, the only decision to be made is whether

0.0

1.0

1.0

One-

dimensional

Attractive

Indifferent Must-be

A

B

C

D E

F

Better

Worse

Figure 17

Figure 18

Distribution of response XY pair Location on the graph of the point

All Attractive 0,1 top left cornerAll One-dimension 1,1 top right cornerEvenly split between Attractive .5,1 middle of top edge, halfway betweenand One-dimensional attractive and one-dimensional-point A

All Must-be 1,0 bottom right cornerEvenly split between One- 1,5 middle of right edge, halfway betweendimensional and Must-be one-dimensional and must-be-point BAll indifferent 0,0 bottom left corner

Evenly split between Must-be .5,0 middle of bottom edge, halfway betweenand Indifferent must-be and indifferent-point CEvenly split between Indifferent 0,.5 middle of left edge, halfway betweenand Attractive indifferent and attractive-point D

Evenly split among Attractive, .5,.5 exact middle of graph-point EOne-dimensional, Must-be, andIndifferentEvenly split between Attractive .5,.5 exact middle of graph, halfway betweenand Must-be attractive and must-be, without an influence

of one-dimensional or independent-point E

Evenly split among attractive, .67,.67 equally spaced between attractive and must-be,one dimensional, and must-be but influenced by one-dimensional-point F

Description of Points in Figure 17

32 Notice the parallels between this graph and the similar graph inSection IIIc (figure 13).33 In our work we actually multiplied by the unscaled Self-statedImportance rating. I have scaled this rating into the interval 0 to 1.0here to avoid redrawing figure 15 on a new scale and then having toexplain the meaning of that scale. To scale the 1 to 9 ratingstypically used in a Self-stated Importance survey into the interval 0to 1.0, subtract 1 from each rating on the 1 to 9 scale and then divideby 8.

1,-5

20 Fall 1993

to include the feature or not. Based on the data,we went back and decided which features weremost worth including. In general, one wants toinclude customer requirements with higher posi-tive Better scores because they add the most tocustomer satisfaction, and customer require-ments with higher (more negative) Worse scoresbecause they prevent the most customer dissatis-faction.

Second, we compiled total grades for ourbest solution versus the competition. If our solu-tion had a feature that the competition didn’t,then we increased our relative score by theamount of Better for that feature; if our solutiondidn’t have a feature that the competition did,then we decreased our relative score by theamount of Worse for that feature. If both we andthe competition had a feature, or neither we northe competition had a feature, our relative scorewas unaffected.

In summary, I came up with a pair of formu-las that can be used to calculate two numbers thatappear to be useful in evaluating Kano question-naire results. We used these numbers in both thesolution generation and concept selection phasesof our voice-of-the-customer efforts.

While I think that this method has merit, Ialso have several reservations:

1.The primary usefulness of the resultingnumbers is in their relative values; that is,in prioritizing the customer requirements.Because the resolution of the Kano data isso low, however, I suspect that the statisti-cal uncertainty of the averages is quite large(tens of percent). Don’t count on thesenumbers to differentiate between close re-sults.

2.This analysis is still based on assigning asingle category (e.g., Attractive, Must-be)to the Kano questionnaire results. Whyshould Attractive be the single result forthree different combinations of answers(like/must-be, like/neutral, like/live-with)?I see this as a basic problem with Kanotheory, but I ignored it.34

3.For some people, it seems redundant toboth multiply by the Self-stated Importancenumber and include the Indifferent cat-egory in the denominator of the Better andWorse calculations.

4.To be more statistically “correct,” oneshould perhaps sum the Self-stated Impor-tance numbers in the appropriate Kano di-mension for each response instead ofsumming the number of instances of a di-

mension and multiplying by the Self-statedImportance average at the end. However, Ididn’t want to go back and rework the data,so I assumed that it would average out.

IVc. Thoughts on Graphical andContinuous Analysis

[Editor’s Note: This section is based on ideas pre-sented by William DuMouchel at a CQM ResearchCommittee meeting on September 10, 1991]

1. Establish the basic plot.The form of analysis described in this section as-sumes there are Q pairs of questions, j = 1, ..., Q,and N respondents, i = 1, ..., N. It also assumesthat a Self-stated Importance questionnaire mayhave been used in parallel with the Kano ques-tionnaire. Thus, there may be three scores foreach potential customer requirement being inves-tigated—Functional, Dysfunctional, and Impor-tance. These three scores are coded as follows:

Functional: Yij =-2 (Dislike), -1 (Live with),0 (Neutral), 2 (Must-be), 4 (Like)

Dysfunctional: Xij =-2 (Like), -1 (Must be),0 (Neutral), 2 (Live with), 4 (Dislike)

Importance: Wij =1 (Not at all Important), ...,9 (Extremely Important).

Thus, if respondent number 8 answered forpotential customer requirement number 6 that heliked good gas mileage, could live with bad gasmileage, and felt the issue of gas mileage was ofmedium importance, then

Y8, 6 = 4,

X8,6 = 2, and

W8,6 = 5

Note that X and Y take on the values -2, -1,0, 2, 4 only. The logic for the asymmetrical scale(beginning from -2, rather than -4) is that Must-be and One-dimensional are stronger responsesthan Reverse or Questionable. Therefore, thescaling should give less weight to the less strongresponses to diminish their influence on the av-erage. The Reverse-type responses are givenless weight by being pulled toward 0.35

34 This issue appears to be at least partially addressed by the ideasdiscussed in subsection IVc on this page.35 It is assumed that only a few of the people responding to thequestionnaire got a Reverse score for the requirement underconsideration. If most of those responding got a Reverse score, theanswers to that requirement could be rescored as described insubsection IIa; see also the discussion of Reverses in subsection Vc.

Fall 1993 21

Figure 19 (below) should help to clarify thepositioning of the various dimensions.

Compare this figure to the Kano EvaluationTable (figure 4) to understand the parallels be-tween the two tables. In particular, note that theinteriors of the two tables (the Rs, Is, Ms, etc.)are laid out in identical fashion. In the methoddescribed in section I, the answers to the func-tional and dysfunctional questions (e.g, like/like,must-be/neutral) are used to look up a single“score” (e.g., R, I) in the Kano Evaluation Table.In the method being described in this section, theanswers to the functional and dysfunctionalquestions are used to specify a point in a two-di-mensional coordinate system.

The most pure or prototypical representa-tions of the Reverse, Indifferent, One-dimen-sional, Must-be, and Attractive points in thistwo-dimensional coordinate system categoriesare identified with the points, respectively:

Reverse X = -2, Y = -2Indifferent: X = 0, Y = 0One-dimensional X = 4, Y = 4Must-be X = 4, Y = 0Attractive: X = 0, Y = 4

These points are underlined and in bolder printin figure 19. All the other combinations of XYpoints appear as interpolations between thesepoints.

Now, for all questions, j = 1, ..., Q, computethe averages of the X (dysfunctional) and Y

(functional) answers across all respondents:

Xave j[ ] =Xij

i∑N

and Yave j[ ] =Yij

i∑N

Then plot the Q points Xave j[ ],Yave j[ ]( ) anduse the number j as the plotting symbol, so youcan identify which question each point repre-sents.

The averages should mostly fall in the range0 to 4, since negative values are eitherQuestionables or Reverses. Questionables willnot be included in the averages. Reverses maybe transformed out of the Reverse category byreversal of the sense of the functional and dys-functional questions over all respondents. Oth-erwise, as described above, there will likely notbe enough Reverses to pull the average negative.

In figure 20 (below) the square where Xaveand Yave range from 0 to 4 is naturally dividedinto quadrants, with the prototypical Attractive,One-dimensional, Must-be, and Indifferentpoints considered to be at the four corners. Thissquare comes from the upper right part of figure19.36

Plots of Average Functionality and AverageDysfunctionality Points for Question J

Figure 19

Functional (-2 to 4) vs. Dysfunctional (-2 to 4)

Bill DuMoucheloriginallypresented thematerial in thissection tomembers of theCQM ResearchCommittee whilehe was employedby Bolt Beranekand Newman Inc.He is now anindependentstatisticalconsultant basedin Belmont, Mass.

36 Note the similarity of this figure to figure 17 in subsection IVb.

Figure 20

Dysfunctional

5

2

1

3

64 7

8

10

9

Attractive One-dimensional

Indifferent Must-be

0 1 2 3 4

Xave

(dysfunctional)

Yave(functional)

1

2

3

4

I I

I

I I

I

I

I

I

O

R

M

A

R RR

R

R

M

M

AAQ

Q

R

Like 4

Must-be 2

Neutral 0

Live with - 1

Dislike -2

Like

-2

Mus

t-be

-1

Neu

tral

0

Live

with

2

Dis

like

4

Dysfunctional

Fun

ctio

nal

22 Fall 1993

When the pairs of coordinates representingthe average responses to each of the Kano ques-tions have been plotted on the grid as shown infigure 20, the nature of each requirement isclearly delineated by the quadrant into whichthat point falls. For instance, a requirement suchas number 5, which falls into the upper left quad-rant, should be viewed as an Attractive element.The closer a point falls to one of the four labeledcorners (the prototypes), the more unanimous thesurvey respondents must have been in theirviews. Conversely, a point such as number 9,which falls near the center of the diagram, is afuzzier result which indicates disagreementamong respondents.

The above is the basic plot. Following are afew possible refinements.

2. Show within question variation

For the jth question, compute

Xstdv[ j]= Standard deviation of the Nscores X

ij,

Ystdv[ j] = Standard deviation of the Nscores Y

ij, and

XYcor[ j]= Correlation coefficient of the Npairs (X

ij,Y

ij.)

Then, on the basic averages graph describedin the previous subsection, plot horizontal errorbars (Xave[ j]±Xstdv[ j]) and vertical error bars(Yave[ j]±Ystdv[ j]) around each point (Xave[ j],Yave[ j]).

A fancier version of this is to use the correla-tion as well as the standard deviation and, in-stead of error bars, plot an ellipse representingthe distribution of XY values. The equation ofthe ellipse is:

If the correlation is zero, the ellipse should passthrough where the ends of the error bars wouldbe. In any case, it should enclose the central 40-50 percent or so of the N points on which it isbased.

(X,Y)X − Xave

Xstd

2

− 2∗ XYcor∗ X − Xave

Xstd∗ Y − Yave

Ystd+ Y − Yave

Ystd

2

= 1 − XYcor2

(X,Y)X − Xave

Xstd

2

− 2∗ XYcor∗ X − Xave

Xstd∗ Y − Yave

Ystd+ Y − Yave

Ystd

2

= 1 − XYcor2

2∗ XYcor∗ X − Xave

Xstd∗ Y − Yave

Ystd+ Y − Yave

Ystd

2

= 1 − XYcor2

3. Show the respondents’ average importanceassigned to each question

From the Self-stated Importance survey that wasdone in parallel with the Kano survey, compute

the average importance,Wave j[ ] =Wij

i∑N

, of

each question and represent its value near theplotted points (Xave j[ ],Yave j)[ ], as in figure 21(next page). For example, draw a filled circlewith radius proportional to Wave[ j] , so thearea of the jth circle would be proportional toWave j[ ]. Alternatively, plot the values of j usedto identify the questions in different colors orshades of gray, representing the range of low tohigh values of Wave .

There is no hard-and-fast way to interpret thediagram in figure 21 for development priorities.The best approach might vary with the number ofpoints falling into each quadrant, with the clus-tering of the points within a quadrant, or with thedegree of differentiation of importance levelswithin a quadrant. For instance, in figure 21there is only one Attractive element. Althoughthat element ranks only as medium in impor-tance, the team might believe that the productneeds a differentiating characteristic.

What makes the most sense is for your groupto view the grid (perhaps without the pointsnumbered, in order to maintain objectivity aboutwhich requirements should be pursued), and thenagree on a decision rule that will work for yourdata.

4. Weight the means for each question accordingto how important that question is rated by eachrespondent.

Instead of plotting Xave,Yave( ) defined asabove, compute

Xwave j[ ] =Wij * Xij

i∑Wij

i∑ and

Ywave j[ ] =Wij * Yij

i∑Wij

i∑and plot points (Xwave j[ ],Ywave j[ ]) as before.This gives those respondents who think a ques-tion is important more voice in deciding whetherthe question is One-dimensional, Must-be, At-tractive, and so on. You should use weightedversions of standard deviations and correlationcoefficients if you want to add error bars or el-lipses to this graph.

Fall 1993 23

Figure 21

Plots of Average Functionality and Average Dysfunctionality Points for Question J with Average Importance Indi-cated for Each Point

Yave(functional)

5

2

1

3

64 7

8

10

9

Attractive One-dimensional

Indifferent Must-be

0 1 2 3 4

Xave(dysfunctional)

3.0-3.94.0-4.95.0-5.96.0-6.97.0-7.9

4

3

2

1

Average importanceintervals:

5. Overlay the basic plot for different groups ofrespondents

Compute the points Xave,Yave( ) separately fortwo or three subsets of the respondents, groupedby background information (demographic data).Then plot the points for the different groups indifferent colors. For example, a red “15” wouldmark the average answer to question 15 for onegroup, and a green “15” would mark the averageanswer to question 15 for another group. Toavoid clutter here you might omit the extra infor-

mation on variation or importance suggested inparts 2 and 3 above.

Va. Theoretical Parallel’s betweenKano’s and Herzberg’s Theories

[Editor’s note: Christopher Bolster has consid-ered Kano’s theories in comparison with thetheories of Herzberg which Kano studied.]

Kano analysis helps us understand the relation-ship between the fulfillment (or nonfulfillment)

24 Fall 1993

Figure 23

of a requirement and the satisfaction or dissatis-faction experienced by the customer. Workingwith social science theories on satisfaction de-veloped by Frederick Herzberg, Kano concludedthat the relationship between fulfillment of aneed and the satisfaction or dissatisfaction expe-rienced is not necessarily linear. He found thathe could sort requirements into distinct classes,and that each class would exhibit a different rela-tionship with respect to satisfaction. Accordingto the Kano’s model, a product induces variousdistinct types of satisfaction or dissatisfaction,depending on whether certain customer needsare completely fulfilled, are only partially met,or go unserved. Kano’s theory, originally termedthe “M-H Property of Quality,” was first pro-posed in a paper published in 1979.37 The basisof knowledge about Kano’s theories within theCQM comes from a subsequent article, entitled“Attractive Quality and Must-be Quality,”38 andfrom teaching by Shoji Shiba.

As mentioned above, the model underlyingKano’s theory has its roots in social psychologyand Motivation-Hygiene theory (M-H theory)39,developed in 1959 by Frederick Herzberg.Herzberg created the theory to explain the wayemployees feel about their work.

Herzberg expounds on his theory at lengthand provides much supporting evidence in hisbook Work and the Nature of Man,40 Herzbergobserves that the set of factors that produce jobsatisfaction are separate and distinct from the setof factors that produce job dissatisfaction. In ef-fect there are two independent axes (figure 22,below).

How Motivational/Hygiene Factorscorrelate to Job Satisfaction

Herzberg thinks of the one axis as the moti-vator axis. In this dimension the employee seekspersonal growth from the tasks being performed;however, absence of this growth does not causepain. The other axis is the hygiene axis. In this

Figure 22

Mental health

Absence of mental health

Absence of mental illness

Mental illness

37 Noriaki Kano, “On M-H Property of Quality,” Nippon QC Gakka,9th Annual Presentation Meeting, Abstracts, pp. 21-26, 1979.38 Noriaki Kano et al., “Attractive Quality and Must-be Quality,”research summary of a presentation given at Nippon QC Gakka:12th Annual Meeting (1982), January 18, 1984.39 Hence the name Kano gave his theory in his 1979 paper.40 Cleveland, World Publishing Co., 1966; see particularly chapter6, pp.71-91. This book is actually the third in a trilogy on thissubject. See also “One more time: How do you motivate employ-ees.” Frederick Herzberg, Harvard Business Review, January-February 1968.

dimension the employee tries to avoid pain fromthe environment; however, avoidance of thispain or the environmental issues that cause thepain does not produce satisfaction.

Some of the issues reported in Herzberg’s re-search as causing increasing levels of satisfac-tion are: growth, advancement, responsibility,work itself, recognition (of work), achievement.Some of the issues cited as causing increasinglevels of dissatisfaction are security, status, rela-tionship with subordinates, personal life, rela-tionship with peers, salary, work conditions,relationship with supervisor, supervision, andcompany policy and administration.

Herzberg goes on to generalize job attitudesto mental health which he notes are similarly ondistinct, parallel lines.

Herzberg’s depiction of Mental Health

Chris Bolster is arecent graduate ofthe MIT SloanSchool ofManagement,where he wasteaching assistantto Dr. ShojiShiba. He spentlast summerworking on a fieldstudy of ConceptEngineering atCQM companiesand helpedrewrite theConceptEngineeringUser’s Guide.

Absence of jobsatisfaction

Job satisfaction

Job satisfaction

Absence of job satisfaction

Hyg

iene

Mot

ivat

or

Fall 1993 25

At the time he wrote his book, Herzbergnoted that many clinicians treated patients withmental health problems as if resolving the con-flicts resulting mental illness would bring mentalhealth. This approach results in doing the hy-giene work to avoid environmental issues butdoes not address psychological growth and self-fulfillment.

Clearly, Kano and M-H theory share somekey features:

• Kano’s satisfaction axis is analogous toHerzberg’s mental health axis. Both extendfrom the origin (a neutral feeling about theproduct or job) through various gradationsof satisfaction. In Kano’s theory, the factorsthat produce only satisfaction are called at-tractive elements. The corresponding ele-ment in Herzberg’s theory is the motivator,sometimes called a “satisfier.”

• Kano’s dissatisfaction axis is analogous toHerzberg’s mental illness axis. Both ex-tend from the origin through various grada-tions of dissatisfaction. In Kano’s theory,the factors that, through their absence, cre-ate dissatisfaction are called Must-be ele-ments. The corresponding element inHerzberg’s theory is the hygiene factor,sometimes referred to as a “dissatisfier.”

In both Kano and M-H theory, there is anasymmetry between the positive and negativeaxes on which the above factors are scaled. Dis-satisfaction is not the opposite of satisfaction.For instance, fulfilling a hygiene or Must-be re-quirement does not lead to satisfaction—in thehealthy individual—because hygiene factors arethose which, when fulfilled, help only to avoiddissatisfaction.

Kano’s theory also, however, includes a thirdclass of requirements that behave as if the posi-tive and negative axes were continuous: one-di-mensional factors. These requirements can causereactions ranging from dissatisfaction, throughindifference, to satisfaction, depending on howwell they are fulfilled.

The one-dimensional element appears tohave no direct analog in M-H theory. Herzbergstates repeatedly that the axes of mental healthand mental illness are distinct. Although heidentifies cases where hygiene factors actuallyprovide satisfaction, this occurs only in “malad-justed” individuals.41 Likewise, highly “growth-oriented” individuals can actually experiencedissatisfaction when motivating factors are notpresent.42

Herzberg notes that the hygiene satisfactionsexperienced by maladjusted individuals areshort-lived and similar in character to opiates.43

The corollary in Kano theory might be that aproduct designed to satisfy only “must-be” re-quirements would not satisfy mainstream con-sumers. Herzberg also states that the unhap-piness experienced by highly growth-orientedindividuals at the lack of motivator factors isqualitatively different from that they experienceat the lack of hygiene factors.44 This is anotherway of saying that the satisfaction and dissatis-faction axes are distinct (satisfaction and dissat-isfaction are not simple opposites). That is alsotrue in Kano theory.

Vb. Wording of the Kano QuestionnaireThe order of the five answers to the pairs ofKano questions seems odd to many people.

1 I like it that way.2. It must be that way.3. I am neutral.4. I can live with it that way.5. I dislike it that way.The question most frequently posed is: Why

is “I like it that way” a stronger statement of sat-isfaction than “It must be that way?” Considerthese responses in the context of the functionalquestion in a pair of questions on a Kano ques-tionnaire. The thought behind the ordering isthat the first response signifies a type of positivesatisfaction, while the last relates to avoidance ofdispleasure. In other words, the logic behind thearrangement of these responses has to do withthe level of pleasure experienced by the cus-tomer. A scale of pleasure is known as a hedonicscale.

An alternative wording of the answers thatseems to differentiate better among the re-sponses is:

1. I enjoy it that way.2. It is a basic necessity or I expect it that way.3. I am neutral.4. I dislike it, but I can live with it that way.5. I dislike it, and I can’t accept it.Some Kano surveys have used this wording.

However, while this alternative wording clarifiesthe difference between “I like it” and “It mustbe,” it may confuse the answers to customer re-quirements that are straight forwardly one-di-

41 Herzberg, p 84.42 Ibid.43 Ibid.44 Ibid.

26 Fall 1993

mensional, such as gas mileage. In the one-di-mensional case, the “I like it” and “I dislike it”pair of answers may be more appropriate that “Ienjoy it” and “I can’t accept it”.

The BBN RS/1 5.0 team45 felt that the CQM-recommended wording of the survey answerswas confusing, and they decided to deviate fromit. They thought phrases such as “I like it thatway” were too ambiguous and they used the fol-lowing wording instead:

1. This would be very helpful to me.2. This is a basic requirement for me.3. This would not affect me.4. This would be a minor inconvenience.5. This would be a major problem for me.

Fred Pouliot of Analog Devices (see subsectionVc) likes a set of wordings he heard from a par-ticipant (name unknown) in a session of theCQM six-day course:

1. I like it.2. I expect it.3. I’m neutral.4. I can tolerate it.5. I dislike it.

This wording appears to distinguish between“like” and “must be” (“expect”) without becom-ing too extravagant (“enjoy”).

Professor Asbjørn Aune of the NorwegianTechnical University (in Trondheim, Norway)has drawn an annotated version of a Kano dia-gram (figure 24, below) illustrating some of theways to think about the Attractive, One-dimen-sional, and Must-be curves and what they maymean in different situations.46

Prof. Aune’s version of the Kano Diagram

Professor Aune’s figure both illustrates someof the difficulty in the wording of the standardanswers to the functional and dysfunctionalquestions, and may suggest some wording for thestandard answers that will help makes thingsmore clear for the person responding to a Kanoquestionnaire. Slightly paraphrasing ProfessorAune, “The One-dimensional class consists ofwhat customers say they want. The Must-beclass can be considered expected quality; it con-sists of expectations that customers do not ver-balize because they assume them to be evident.Attractive is the class of exciting quality; cus-tomers do not expect the quality characteristicsin this class, but they recognize them as im-provements and like them.”

Reinhart Richter of BBN suggested that an-swers that don’t make sense for a particular pairof questions should perhaps be left off in the in-terests of clarity. It might also be useful tochange the wording of the standard answers tomake them better match certain questions, or atleast to change the standard wording dependingon the general context of a particular question-naire. Such changes in wording could be danger-ous if they are done casually, leading toinconsistent or biased answers. On the otherhand, it would be useful to have the results of asystematic effort to see which wording of thestandard answers works best and whetherchanges from question to question improve theusefulness of Kano questionnaires or bias the an-swers.

Given the unfamiliarity customers have withKano surveys and the confusion that customersseem to have with the answers to the question, itis especially important to provide good instruc-tions for answering a Kano survey. A sampleKano questionnaire, based largely on instruc-tions provided by Charles Berger of BBN Soft-ware Products, is shown on the facing page.

One might consider expanding Berger’squestionnaire with as follows:

• Do not be distracted with the order of thestandard answers. Simply select the answerthat seems most appropriate to you for thequestion being answered.

• Do not be distracted by answers that don’tseem to apply. Simply select the most ap-propriate answer that does apply.

Figure 24

Must-be

Dysfunctional

Love the Product

Hate the Product

FullyFunctional

Attractive

Surprises–not expressed

–customer tailored-transcendent

Performance–stated

–specified–measurable

–technical

Basic Requirements–implied

–self-evident–not mentioned

–taken for granted

One-dimensional

45 See section III.46 Seminar Proceedings, “World Class Quality—the Role of TopManagement,” the Norwegian Academy of Technological Sciences,November 7-8, 1991, Oslo, Norway, pp. 13-14.

Fall 1993 27

1a. If the eggs are served hot, how do you feel?

1b. If the eggs are not served hot, how do you feel?

2a. If the eggs are offered in a variety of ways (e.g. poached, fried, scrambled, hard-boiled, omelet, soft- boiled), how do you feel?

2b. If the eggs are offered as only fried or scrambled, how do you feel?

1. I like it that way.2. It is a basic necessity or I expect it that way3. I am neutral.4. I dislike it but can except it.5. I dislike and can except it.




Circle the number in front of the option that best describes how you feel, for each question.

"Kano Survey"

This questionnaire asks pairs of multiple-choice questions about potential product capabilities. Half of each pair of questions asks how you would feel if the product we hope to supply in the future included a particular capability; the other half of each pair asks how you would feel if the capability were not provided.

Example Instructions on How to Fill Out the Kano Survey

The breakfast menu example shown here demonstrates how one customer might answer such pairs of questions. In this example, question 1a (the eggs are served hot) asks how the customer would feel if a capability or feature were provided, and the customer's answer is circled, while question 1b (the eggs are not served hot) asks how the customer would feel if a capability were somehow limited or absent and the customer's answer is circled. The second question, 2a, asks how the customer would feel if there were a variety of cooking options for the eggs (poached, fried, scrambled, hard-boiled, omelet, etc.), and question 2b asks how the customer would feel if eggs were offered only as fried or scrambled. From collecting responses from many customers we can derive detailed information about customers' needs.

Charles Berger isan employee ofBBN’s SoftwareProductsDivision, and hasused Kano’smethods in thespecification ofproducts.

28 Fall 1993

• The pairs of questions will be used to clas-sify each customer requirement so we candecide how best to handle it. The pairs arequestions are not used to rank-order thecustomer requirements. Therefore, do nottry to “second guess” the survey in order toindicate higher priority for the customer re-quirements you care about most.

• The first answer to each question applieswhen the question is about something thatwould mean at least a little something spe-cial to the respondent. The second answerapplies when the question is about some-thing that the respondent takes for grantedshould be in the product.

Vc. Theoretical Issues of Kano’s Methods[Editor’s note: Fred Pouliot has considered thetheoretical details of Kano’s methods. His sug-gestions for the wording of the five questionswere given in section Vb, and he uses this word-ing in his theoretical analysis.]

It is possible to derive the curves in the Kanodiagram from the Kano Evaluation Table as fol-lows.47 Plot the axes of the Kano Diagram andlabel vertical levels of the graph with the word-ings of the answers to questions on a Kano ques-tionnaire, as in figure 25 (below).

In figure 25, the “Must-be” level is only alittle above neutral because Must-be is only a

weak statement of satisfaction—it is more astatement of lack of dissatisfaction, though cer-tainly more positive than neutral. Symmetri-cally, “can live with” is not a strong statement ofdissatisfaction, but its grudging acceptance ismore negative than neutral.

It is now possible to plot lines representingall 25 combinations of answers represented inthe Kano Evaluation Table. For instance, thelines represented by the following three pairs ofend points

(functional, like); (dysfunctional, must be)(functional, like); (dysfunctional, neutral)(functional, like); (dysfunctional, can live with)

are shown in figure 26 on the facing page (andlabeled by their cell numbers in the Kano Evalu-ation Table—reproduced for the reader’s conve-nience in figure 27 on the facing page). Fromthese three lines one can roughly envisionKano’s Attractive curve which is shown in boldin figure 26.

The other 22 lines implicit in the KanoEvaluation Table (figure 27) can be plotted simi-larly and the general shapes of the rest of theKano curves can be imagined from these lines.

But there is a more systematic and, I believe,more well-founded derivation of the Kano

Figure 25

Plotting the Kano Answers on the Kano Diagram Axes

Satisfied

Dissatisfied

ikL e

Must-be

Neutral

Can live with

Dislike

Dysfunctional Fully Functional

47 Dave Walden of BBN circulated informal notes on this simplederivation in May 1991.

Fall 1993 29

Plot of the three Attractive cells from theKano Evaluation Table

Figure 26


Func-tional

1. like

2. must-be

3. neutral

4. live with

5. dislike

Q

R

R

R

R

A

I

I

I

R

A A

I I

II

II

R R

O

M

M

M

Q

Dysfunctional1.

like2.

must-be3.

neutral4.

live with5.

dislike

Customer Requirement is:A: AttractiveM: Must-beR: Reverse

O: One-dimensionalQ: Questionable I: Indifferent


Figure 27

curves. What follows is an explanation of thisderivation, together with some observations onhow to improve the use of Kano’s methods.48

The shape of the Kano curves as usually pre-sented (see figure 28, next page) is somewhat in-consistent with the Kano Evaluation Table usedto translate pairs of answers to the questions (seefigure 27). This is clear from figure 26 where wehad to stretch to image the Attractive curve as

being representative of the three lines derivedfrom the Kano Evaluation Table.

This inconsistency has led to some confu-sion. For example, consider the Attractive curvein Figure 28. It starts above the X axis on thedysfunctional side, and then proceeds asymptoti-

48 Fred Pouliot’s ideas have been circulating informally in the CQMsince late 1991.- Ed.

Fred Pouliot isdivision managerof the IC TestSystem Divisionof Analog DevicesInc. He hasfrequentlyvolunteered as aCQM 6-DayCourse leaderand is very muchinvolved withTQM training atAnalog.

imaginedAttractive curve

Must-be

Neutral FullyFunctional

Dys-functional

Satisfied

Dissatisfied

3 Attractive lines

1-2

1-3

1-4 Can live with

Dislike

Like

30 Fall 1993

Kano Diagram

Figure 28

Attractive, Must-be, and One-dimensional Curves

Figure 29

A ttractive

Must-b e

One-dimensional

Dysfunctional

Satisfaction (Delight)

Dissatisfaction

Functional

Fall 1993 31

cally to Delight on the functional side. Thecurve consists of all positive values. This is in-consistent with the dysfunctional/live with as-pect of cell 1-4 in the Kano Evaluation Table(Figure 27). There is a similar inconsistencywith the Must-be curve. The curve comprisesall negative values, yet, one cell, 2-5, implies apositive value (functional/must-be).

Figure 29 (facing page) illustrates a detailedderivation of the Attractive, Must-be and One-dimensional curves. The data points (big blackdots) are plotted from the Kano Evaluation Tableand are labeled with functional-dysfunctionalnumber pairs from that table, but there are someassumptions that make the derivation in figure29 work well.

• Assumption 1: The word “like” is meant tobe a very strong like, which in the limit be-comes complete customer “delight.” It cantherefore be considered to be a point thatexists at +infinity. A point close to the topof the Y axis is shown to get across the con-cept of rapid improvement in satisfaction asfunctionality increases; that is, the conceptthat an exponential relationship exists.

• Assumption 2: The word “dislike” is asstrong in the negative direction as “like” isin the positive direction, i.e., severe dissat-isfaction. As with “like”, the curves for“dislike” coincide at a point close to, andnear the bottom of, the Y axis.

• Assumption 3: A relatively small differ-ence is represented between the terms“must-be” and “neutral”. Similarly there is

Figure 30

a small difference between “can live with”and “Neutral”.

Cells 2-2 and 4-4 of the Kano EvaluationTable are also logically inconsistent; for ex-ample, a requirement that is rated as Must-befunctional cannot simultaneously be rated asMust-be dysfunctional. I suggest that cells 2-2and 4-4 in the Kano Evaluation Table bechanged from “I” to “Q.”

The points discussed thus far in my detailedderivation are shown in the partially completedKano Evaluation Table in figure 30, along withfundamental combinations for Indifference, “I.”

The four Indifferent points that have been in-cluded in figure 30 are straight lines shown inbold in figure 31 (next page).

The lines for “Q” derived from the KanoEvaluation Table in Figure 24 are shown in boldin Figure 32 (next page).

The upper right half of the Kano EvaluationTable can be viewed as the positive half, since allcombinations are the result of customer re-sponses being more positive for the Functionalquestion than for the Dysfunctional question.There is symmetry about the diagonal as shownin figure 33 (page 63).

Each element on the positive half has a cor-responding element on the negative side that isthe exact opposite of the original. These are“Reversals,” i.e., for each Kano dimension, K

m,n,

there is a reversal dimension, Rn,m

. The table atthe end of this paragraph shows the originalKano dimension on the positive half (above maindiagonal) of the matrix, its corresponding rever-

Partially Derived Kano Evaluation Table

32 Fall 1993

Questionable Lines

Figure 31

Neutral Lines

Delight

Severe Dissatisfaction

FullyFunctional

FullyDysfunctional

Must-be

Neutral

Live With

Like

Dislike

2-3

2-43-3

3-4

Figure 32

Delight


FullyFunctional

FullyDysfunctional

Must Be

Neutral

Live With

Like

Dislike

Q

Q

Q

Q

Fall 1993 33

Partially Derived Kano Evaluation Table with Diagonal Shown

Dysfunctional1. 2. 3. 4. 5.

I like it I expect it I'm neutral I can tolerate it I dislike it1. I like it Q A A A O2. I expect it Q I I M

Funct- 3. I'm neutral I I M

ional 4. I can tolerate it Q M5. I dislike it Q

Requirement is:There is symmetry about this diagonal

A: Attractive O: One-dimensionalM: Must-be Q: Questionable resultR: Reverse I: Indifferent

Figure 33

sal element (below main diagonal), and the sub-script of the reversal element.

Location of Location ofKano Dimension Reversal Reversalon (+) Half on (-) Half Dimension

1-2 2-1 RA

1-3 3-1 RA

1-4 4-1 RA

1-5 5-1 RO

2-3 3-2 RI

2-4 4-2 RI

3-4 4-3 RI

2-5 5-2 RM

3-5 5-3 RM

4-5 5-4 RM

It can be argued that the reversal of an Indif-ferent is still an Indifferent. Therefore, nothing islost by leaving cells 3-2, 4-2, and 4-3 as Indiffer-ent. The Kano Evaluation Table appropriatelyextended is shown in figure 34 (below).

The reversals result in mirror images of theoriginal Kano curves, as shown by the dashedcurves in Figure 35 (next page).

As we now complete the derivation of the re-vised Kano diagram, we can ignore all of the Qlines in figure 32 and figure 35, since they are forthe most part the result of poorly phrased ques-tions or poorly developed requirements.Through the concept engineering process, for in-stance, we strive to eliminate or minimize theiroccurrence.

Figure 34

Requirement is:A: Attractive O: One-dimensionalM: Must-be Q: QuestionableR: Reverse I: Indifferent

Kano Evaluation Table with Reversals Shown

Dysfunctional

Fun

ctio

nal

1. 2. 3. 4. 5.

1. I like it

2. I expect it

3. I'm neutral

4. I can tolerate it

5. I dislike it

I like it I expect it I'm neutral I can tolerate it I dislike it

Q A A A O

R Q I I M

I I I M

I I Q M

Q

A

RA

RA

RO

RM

RM

RM

34 Fall 1993

Cells in Upper Right Portion of Kano Evaluation Tableand their Reversals

Figure 36

Delight


FullyFunctional

FullyDysfunctional

Must-be

Neutral

Live with

Like

Dislike

Attractive

One-dimensional

Must-be

“Average” Attractive, Must-be, and One-dimensional Curves

Figure 35


Delight

FullyFunctional

FullyDys-

functional

Neutral

Live With

Like

Dislike

Must-be

Fall 1993 35

Figure 37

Degenerate Indifferent Line

FullyFunctional

FullyDysfunctional

Delight

SevereDissatisfaction

Must-beIndifferent

One-dimensional

Attractive

Next, averaging the set of curves for Attrac-tive, and Must-be, in figure 34 results in the dia-gram shown in figure 36, on the facing page (theIndifferent lines haven’t yet been averaged):

To simplify the Indifferent curves we can goback to assumption 1 on page 31, and considerthe difference between Like and Must-be to beso great as to render the distance between the In-different lines to be near zero. In the limit thecurves for Indifferent become the abscissa; thatis, the curve for Indifferent is a straight line ex-isting along the X axis.

Thus, we have derived the Modified KanoDiagram shown in figure 37, above (without re-versal items). Note: The Attractive and Must-becurves are coincident with the X axis rather thanbeing asymptotic.

Finally, in figure 38 (next page) we show thecurves from figure 37 with the reversal curves.

From the revised Kano Evaluation Table infigure 34 and figure 38 the reader may deducewhy retroactively reevaluating survey questionsthat resulted in a Reverse rating can produce use-ful data, as if the reverse of those questions hadbeen asked.

36 Fall 1993

Kano Diagram with Reversal Items

Delight

One-dimensionalReversed One-dimensional

FullyDysfunctional


FullyFunctional

Indifferent

Attractive

Must-be

Reversed Attractive

Reversed Must-be

Figure 38

Jour

nal O

n-Li

ne

The Center for Quality of Management Journal is a forum for disseminating theexperience of organizations learning to implement modern management practices. Itseeks to capture experiences and ideas that may be useful to others working to createcustomer-driven, continuously improving organizations.

The CQM Journal is refereed. However, it is not an academic publication. Experiencesand ideas will be published if they seem likely to be useful to others seeking to improvetheir organizations.

Send to:

The Center for Quality of Management JournalEditorial DepartmentOne Alewife Center, Suite 450Cambridge, MA 02140Tel. 617-873-8950 Fax 617-873-8980E-mail: [email protected]

If you have thoughts for a paper and you would like to discuss it with us, please write,call or submit an outline. We welcome your ideas.

Final Manuscript Requirements:

Entire manuscript should be double-spaced, including footnotes, references, etc. Textshould include all the elements listed below. Generally, The CQM Journal follows theeditorial principles of The Chicago Manual of Style. We strongly prefer submissions ineletronic format for all text and as many of the figures as possible. IBM-based software(particularly Microsoft Word for Windows) is preferable to Macintosh-based software ifyou have a choice, but is by no means a requirement.

Please include:

1. Title page, stating the type of article (e.g., 7-Step case study, research paper, shortcommunication, letter to the editor, etc.), main title, subtitle, and authors’ full name(s),affiliation(s), and the address/phone/fax of the submitting author;

2. All needed figures, tables, and photographs (see below);

3. Footnotes (if appropriate), numbered consecutively from the beginning to the end ofthe article;

4. Reference list, if appropriate.

Figures, Tables and Photographs:

If you can, insert each figure or table into the text where you would like it to fall. Figuresshould be composed to conform to one of two widths: 3 1/8 or 6 1/2 inches. Themaximum height for any figure is 9 3/8 inches. Text within figures should not besmaller than 5 points and lines not less than 1/4 point at the figure’s final size. Figuresshould labeled with the figure number underneath and title on top. Be sure that the textmentions each figure or table.

Please retain separate PICT or TIFF files of figures generated in drawing programs and afile with the text only for final submission.

Editorial Board

David Walden, ChairmanCenter for Quality of Management

Stephen GravesProfessor & LFM Co-DirectorMassachusetts Institute of Technology

Ted WallsBoston College

Robert Chapman WoodWriter

Alan GrahamConsultantPugh-Roberts Associates

Shoji ShibaTokiwa University

Production Team

Eric BergemannPublisher

Kevin M. YoungDesign & Production

Jay HowlandCopy Editing

CQM Officers

Ray StataChairman

Gary BurchillPresident

Thomas H. LeeTreasurer and President Emeritus

William WiseClerk

CENTER FOR QUALITY OF MANAGEMENT JOURNAL

center for quality of management journalwalden-family.com/public/cqm-journal/2-4-whole-issue.pdf ·...

Documents