Customer & product
Kano Model
A model of how product attributes relate to customer satisfaction, showing that some qualities merely prevent dissatisfaction, some scale with performance and some delight unexpectedly, so that features can be classified and prioritised by the kind of satisfaction they buy.
Also known as Theory of attractive quality, Kano analysis, Attractive quality model. First set out by Noriaki Kano, with Nobuhiko Seraku, Fumio Takahashi and Shinichi Tsuji in 1984; the primary source is cited in full below.
- Format
- Mapping
- Level
- Product
- Best for
- Understand customers · Prioritise · Evaluate options
- Decision stage
- Diagnose · Explore options · Decide
- Difficulty
- Intermediate
- Time to apply
- Two to four weeks end to end: a day to decompose attributes and draft question pairs, one to two weeks in field, and a few days for classification and the investment read.
Plate · The model
The components
Must-be quality
Attributes customers assume without asking. Their absence causes outrage; their presence earns nothing, because fulfilment is taken as given. They set the floor a product must clear before anything else it does is even noticed.
Signals of strength
Customers mention it only when it fails · Dysfunctional answers cluster on 'dislike'; functional answers on 'must be' or neutral · Complaints and returns cite it; praise never does · Competitors all provide it, unremarked · Investment beyond adequacy produces no measurable satisfaction gain
One-dimensional (performance) quality
Attributes where satisfaction rises roughly in proportion to performance: more is better, less is worse. These are the qualities customers consciously compare and talk about, and where head-to-head competition usually lives.
Signals of strength
Customers can rank options on it and do · Functional answers cluster on 'like'; dysfunctional on 'dislike' · It appears in comparison tables, reviews and buying guides · Willingness to pay moves with the level provided · Both satisfaction and dissatisfaction coefficients are high
Attractive quality
Attributes that delight when present and cost nothing when absent, because nobody expected them. They generate disproportionate satisfaction, word of mouth and differentiation, and they are the model's strategic prize.
Signals of strength
Surprise and delight in reactions; customers show the feature to others · Functional answers cluster on 'like'; dysfunctional answers are neutral or tolerant · Absence generates no complaints at all · Not yet offered by most competitors · Mentioned unprompted in reviews as a reason for recommendation
Indifferent quality
Attributes whose presence or absence customers simply do not register in satisfaction. Every mature product accumulates them, and they quietly consume engineering and marketing spend that the model exists to reclaim.
Signals of strength
Neutral answers dominate both questions · Usage data shows the feature is rarely touched · Nobody mentions it in interviews, reviews or complaints · Removing it in a test provokes no reaction · It survives on internal attachment rather than customer evidence
Reverse quality
Attributes whose presence actively dissatisfies some customers, who prefer the product without them. Reverse findings are segment signals: what one group welcomes as capability another experiences as clutter, complexity or intrusion.
Signals of strength
Functional answers cluster on 'dislike' for a nontrivial share of respondents · Requests to switch it off, hide it or pay less without it · Polarised reviews praising and damning the same attribute · A simpler competitor is winning a subset of your customers · Category flips between segments in the survey data
When it earns its keep
- You are deciding which features to build, keep or cut and want a customer-grounded basis for the trade-offs, rather than a stakeholder shouting match or a flat vote count.
- Satisfaction scores are stagnant despite steady investment, and you suspect spend is going into attributes customers now take for granted.
- You are specifying a new product and need to separate the non-negotiable basics from the differentiators before the budget is carved up.
- Competing proposals promise 'delight' and you want to test, with structured customer evidence, which candidate features would actually excite anyone.
And when it doesn't
- You cannot survey real customers of the relevant segment. The model without the questionnaire is a diagram of opinions, and internal guesses about what delights customers are exactly what it exists to check.
- The question is which underlying customer need to serve, rather than which attributes of a chosen offer to invest in. Jobs to be Done and journey mapping sit upstream of Kano.
- Attributes cannot be meaningfully isolated, as with deeply bundled experiences where customers judge the whole rather than the parts; forced decomposition produces clean categories of doubtful meaning.
- You need the answer to hold for years. Kano himself showed attractive qualities decay into expected ones over time, so a classification is a dated snapshot, not a durable ranking.
How to run it
Before starting, gather the inputs the analysis depends on:
- A defined product or product concept, decomposed into discrete candidate attributes or features, each expressible as present or absent.
- Access to a representative sample of current or target customers for the paired-question survey, ideally segmented, since categories often differ by segment.
- For each attribute, a functional and dysfunctional question pair worded concretely enough that respondents can imagine both states.
- Rough cost or effort estimates per attribute, so the classification can be turned into an investment decision rather than a poster.
- 1
Decompose the offer into attributes
List the candidate features or qualities to classify, each specific enough to imagine present or absent. 'Build quality' is too vague to survey; 'metal rather than plastic gear housing' can be answered.
- 2
Write the paired questions
For each attribute, write a functional question (how would you feel if the product had this?) and a dysfunctional question (how would you feel if it did not?). Respondents answer both on the five-point scale: I like it that way; it must be that way; I am neutral; I can live with it that way; I dislike it that way.
- 3
Survey real customers
Run the questionnaire with a representative sample of the target segment. Keep it short; the paired format is cognitively heavy, and a survey covering forty attributes will be answered carelessly from the fifteenth onward.
- 4
Classify with the evaluation table
Cross-tabulate each respondent's functional and dysfunctional answers using the standard evaluation table, formalised by Berger and colleagues, assigning each attribute to must-be, one-dimensional, attractive, indifferent or reverse, with questionable results flagged for contradictory answers.
- 5
Aggregate and read by segment
Tally categories across respondents, using the modal category or Berger's satisfaction and dissatisfaction coefficients to express strength. Read the results by segment before pooling; an attribute can be attractive to one group and reverse to another, and the average hides the conflict.
- 6
Decide the portfolio and revisit
Fund must-bes to threshold and no further, compete on the one-dimensionals that matter most per unit of cost, select a few affordable attractives, and stop spending on indifferents. Re-run the study periodically, because today's delighter is on its way to being tomorrow's expectation.
Reading the result
A classification of each candidate attribute into must-be, one-dimensional, attractive, indifferent or reverse, backed by paired-question survey data and typically summarised with satisfaction and dissatisfaction coefficients, feeding a feature investment decision.
- Remember the canonical form is a two-axis chart, degree of implementation across the horizontal axis and customer satisfaction on the vertical, on which must-be attributes curve up to a ceiling of mere acceptance, one-dimensionals run diagonally and attractives curve up from neutral. The stacked bands here list the categories; the curves are the argument.
- Read must-bes as a threshold, not a ladder: meet them fully, then stop. Every pound spent exceeding a must-be is a pound unavailable for a one-dimensional or an attractive.
- Treat indifferent classifications as found money, the licence to simplify, and treat reverse findings as segmentation questions before treating them as feature verdicts.
- Date-stamp the results. Categories migrate with expectations, usually downward from attractive toward must-be, so a Kano study describes this market this year.
A worked example
A kitchen appliance manufacturer decides the feature set for a new stand mixer
A Midlands-based kitchen appliance manufacturer is specifying a new mid-range stand mixer to sit between budget imports and premium heritage brands. Engineering has fourteen candidate features and budget for perhaps nine. The team decomposes the concept into attributes, surveys 240 home bakers with paired functional and dysfunctional questions, and classifies the answers with the standard evaluation table. Five findings shape the decision.
- Must-be quality
- A stable, non-walking base and a bowl that locks positively classify overwhelmingly as must-be: dysfunctional answers are 'dislike' almost without exception, functional answers earn no enthusiasm. Nobody will buy the mixer for these; anybody would return it without them. The specification meets them fully and spends nothing more, resisting an engineering proposal for an over-specified base that the data says no customer would ever notice.
- One-dimensional (performance) quality
- Motor power, mixing bowl capacity and quietness classify as one-dimensional: satisfaction scales with the level offered, and these attributes dominate the comparison tables in reviews. This is where the mid-range positioning is won or lost, so the budget concentrates here, targeting class-leading quietness because rivals compete on watts.
- Attractive quality
- An integrated scale in the bowl, weighing ingredients as they are added, classifies as attractive: functional answers cluster on 'like', while its absence troubles nobody because no expectation exists. A second candidate, app connectivity, was expected by the product team to delight and did not. The scale makes the specification as the launch differentiator; connectivity is dropped.
- Indifferent quality
- A retro colour range in twelve finishes and a branded dust cover classify as indifferent: neutral answers dominate both questions and interviews never mention them. Cutting the range to four colours and making the cover an accessory releases tooling and inventory cost that part-funds the integrated scale. This is the model doing its quiet, unglamorous work.
- Reverse quality
- Preset automatic programmes split the sample: convenience-oriented respondents like them, but the keen-baker segment answers 'dislike' on the functional question, reading presets as the machine overruling their judgement. For a mixer aimed at keen bakers this is reverse quality. The presets are cut in favour of a plain speed dial, and the finding is filed as evidence for a possible convenience-segment model later.
The read. The study moves real money: it stops over-investment in must-bes, concentrates spend on the one-dimensionals reviewers measure, funds one genuine attractive by cutting indifferents, and prevents a reverse-quality feature from alienating the core segment. The honest caveats are that the classification describes today's expectations, since an integrated scale will not stay attractive once copied, and that a different segment would have produced a different map. The team diarises a re-run after launch year one.
Pitfalls
- Classifying features by internal debate and drawing the diagram afterwards. The questionnaire is the method; without the paired questions and evaluation table you have opinions arranged in a Japanese chart.
- Wording attributes abstractly, so respondents imagine different things and the answers blur toward questionable or indifferent. Concreteness in the question pairs is most of the craft.
- Pooling segments into one classification. Reverse and attractive findings in particular differ by segment, and the pooled mode can be a category that no actual group of customers holds.
- Reading must-be as unimportant because it scores no delight. Must-bes are the most important attributes in the model; they are simply the ones where the correct investment is exactly enough and no more.
- Treating the study as permanent. Kano's own later work on the life cycle of quality attributes shows delighters decay into expectations, so an unrefreshed classification quietly overstates yesterday's differentiators.
What the critics say
The classification methods themselves are contested. Mikulic and Prebezac review the main techniques and show the paired-question approach and its rivals can assign the same attribute to different categories, with results sensitive to question wording, scale interpretation and the analysis rule chosen, which undermines confidence in any single study's categories.
Mikulic, J. and Prebezac, D. (2011) 'A critical review of techniques for classifying quality attributes in the Kano model', Managing Service Quality, 21(1), pp. 46-66.
Categories are unstable over time and across cultures: two decades of empirical work show attractive attributes migrating to must-be as expectations rise, and category assignments varying between markets, so the model offers a snapshot whose shelf life is short in fast-moving categories.
Lofgren, M. and Witell, L. (2008) 'Two Decades of Using Kano's Theory of Attractive Quality: A Literature Review', Quality Management Journal, 15(1), pp. 59-75.
The stated-response method measures what customers say about hypothetical presence and absence, and self-reports of future delight correlate imperfectly with revealed behaviour and willingness to pay. Even sympathetic methodologists such as Berger and colleagues note the results need corroboration from behavioural evidence before they justify large investments.
Berger, C. et al. (1993) 'Kano's Methods for Understanding Customer-defined Quality', Center for Quality Management Journal, 2(4), pp. 3-36.
Run it as a workshop
This is two working sessions either side of a real customer survey, not a single workshop, and the survey fieldwork is the actual method, not preparation for it. Session 1 needs product, engineering and whoever owns customer research in the room together, since the attribute list and question wording set the ceiling on how useful the data can be. Do not run Session 2 until real survey responses are in hand; a Kano session built on internal guesses about what would delight customers is exactly the failure mode the model exists to catch.
Running order
| Session 1 (half day): decompose the product concept into discrete, concretely worded candidate attributes | 60 min |
| Session 1: draft the functional and dysfunctional question pair for each attribute | 75 min |
| Session 1: agree the target segments, sample size and fielding plan for the survey gap | 30 min |
| Session 2 (half day, held once survey data is back): walk the group through the evaluation table classification for each attribute | 75 min |
| Session 2: read results by segment before pooling, flagging any attribute that splits between segments | 45 min |
| Session 2: decide the feature portfolio, funding must-bes to threshold, competing on one-dimensionals, selecting a few attractives, cutting indifferents | 60 min |
| Session 2: diarise the re-run date and note attributes worth re-testing sooner | 15 min |
| Total | 6h |
You will need
- The candidate attribute list from Session 1, each worded specifically enough for a respondent to imagine present or absent
- The paired functional/dysfunctional question template and the standard five-point response scale
- Segmented survey results, ideally visualised per segment before any pooled view is shown
- The standard Kano evaluation table for live classification, plus satisfaction and dissatisfaction coefficients if the sample supports them
- A portfolio decision sheet mapping each attribute to fund, cut or hold, with rough cost or effort against it
Traps to avoid
- Classifying attributes by team debate and skipping the survey. The paired-question fieldwork is the method itself; a diagram produced from a workshop discussion alone is opinions arranged in a Kano-shaped chart, which the entry's own pitfalls call out directly.
- Wording attributes abstractly in Session 1. Vague attributes produce vague, questionable-category answers; push the room to concreteness before the survey goes out, since it cannot be fixed afterwards.
- Pooling all segments into one number before anyone looks at them separately. Attractive and reverse findings in particular differ by segment, and a pooled average can land on a category no real group of customers actually holds.
- Treating must-be attributes as unimportant because they score no delight in the room's reaction. They are the most important attributes in the model; the correct decision is to fund them to threshold and stop, not to under-invest.
- Filing the classification as permanent. Today's attractive attribute is next year's must-be; leave Session 2 with a re-run date on the calendar, not just a portfolio decision.
Sources and further reading
- Kano, N., Seraku, N., Takahashi, F. and Tsuji, S. (1984) 'Attractive Quality and Must-Be Quality', Hinshitsu: The Journal of the Japanese Society for Quality Control, 14(2), pp. 39-48. ↗
- Berger, C., Blauth, R., Boger, D. et al. (1993) 'Kano's Methods for Understanding Customer-defined Quality', Center for Quality Management Journal, 2(4), pp. 3-36.
- Lofgren, M. and Witell, L. (2008) 'Two Decades of Using Kano's Theory of Attractive Quality: A Literature Review', Quality Management Journal, 15(1), pp. 59-75.
- Mikulic, J. and Prebezac, D. (2011) 'A critical review of techniques for classifying quality attributes in the Kano model', Managing Service Quality, 21(1), pp. 46-66. ↗