The Kano questionnaire: find out which features customers would actually miss
Ask customers whether they want a feature and almost everyone says yes. The Kano questionnaire asks two questions per feature and tells you which features are expected, which ones delight, and which ones nobody would notice.
By SurveyLane · The team building SurveyLane
Ask a customer whether they would like a new feature and you will almost always hear yes. Yes is free and polite. So a survey built on "would you use this?" gives you a roadmap where everything scores 4 or 5. The Kano questionnaire changes one thing: you ask about each feature twice, once as present and once as absent, and the pair of answers tells you what kind of feature it is.
Why "would you use this feature?" tells you so little
A single interest question mixes up three kinds of feature. Some people take for granted: no thanks when present, but they leave when it is missing. Some they want more of. And some nobody asked for, yet people are pleased to find them. All three score high on "how useful would this be?", so the question cannot tell them apart.
That confusion costs money. Postpone a taken-for-granted feature and churn creeps up. Treat a pleasant surprise as a requirement and you spend a quarter polishing something nobody would have missed.
Satisfaction does not move in a straight line
In 1984 Noriaki Kano, Nobuhiko Seraku, Fumio Takahashi and Shin-ichi Tsuji published "Attractive quality and must-be quality" in Hinshitsu, the journal of the Japanese Society for Quality Control. Their point: satisfaction responds differently to different attributes.
Must-be attributes cause dissatisfaction when missing and indifference when present: an export that does not lose data, a password reset that works. One-dimensional attributes satisfy more the better they perform, like speed or accuracy. Attractive attributes please when present and draw no complaints when absent. Indifferent attributes move nothing. Reverse attributes annoy part of your audience when you add them. The post on key driver analysis touches on the same asymmetry.
The question pair that does the work
Every feature gets two questions. The functional one asks how the respondent would feel if the product had the feature. The dysfunctional one asks how they would feel if it did not. Both use the same five answer options, and you classify the pair.
For a survey tool it might read like this. Functional: "If you could schedule a survey to close automatically on a date you choose, how would you feel?" Dysfunctional: "If you could not schedule a survey to close, and had to close it by hand, how would you feel?" "I like it" followed by "I am neutral" means the feature is attractive to that person. "I expect it" followed by "I dislike it" means it is a must-be, and its absence is a problem today.
Most teams have never asked the dysfunctional question, and that is where the information sits. Picturing the product without something takes honest thought. Liking an idea takes none.
Writing the two questions well
Kano studies usually fail on wording. Four rules catch most of it.
Describe the feature as something the respondent can picture. "Smart routing v2" means nothing to a customer. "People who answer no to the first question skip the follow-ups automatically" does. Show a sketch if you have one.
Make the dysfunctional question a true negation of the functional one. "If the export did not include open answers" is the opposite of "if the export included open answers". "If the export were slower" is a different attribute and needs its own pair.
Keep both questions neutral. "How would you feel if we finally added dark mode?" pushes before anyone has answered. The post on writing better survey questions covers leading wording, and here every rule applies twice.
Keep the list short. Fifteen features means thirty near-identical questions and a lot of straight-lining. Keep it to eight to twelve per respondent and split longer lists across random subsets.
The five answer options
The scale most teams use runs: I like it, I expect it, I am neutral, I can tolerate it, I dislike it. Charles Berger and colleagues discussed several wordings in a 1993 special issue of the Center for Quality of Management Journal, which turned Kano's idea into a working method. Pick one and never vary it.
The middle three cause the trouble. "I expect it" and "I can tolerate it" are awkward in some languages, and people sometimes read "I expect it" as "I assumed you already had this". A pretest with five or six people who explain their choice out loud, as in the post on testing a survey before you send it, finds this in an afternoon.
Turning answer pairs into categories
Each respondent's pair maps to one category through the evaluation table from Berger et al. Rows are the functional answer, columns the dysfunctional answer.
| Functional / dysfunctional | Like | Expect | Neutral | Tolerate | Dislike |
|---|---|---|---|---|---|
| Like | Q | A | A | A | O |
| Expect | R | I | I | I | M |
| Neutral | R | I | I | I | M |
| Tolerate | R | I | I | I | M |
| Dislike | R | R | R | R | Q |
A is attractive, O one-dimensional, M must-be, I indifferent, R reverse, Q questionable. Questionable means a contradictory pair, like liking both the presence and the absence of the same feature. If more than a few percent land there, the pair was probably confusing, so fix the wording before trusting that feature's result.
Look at the whole distribution
The quick analysis takes the most frequent category per feature. That hides a lot. A feature with 34 percent attractive, 32 percent indifferent and 30 percent must-be gets labelled attractive, while almost a third of your customers would be annoyed without it.
So read the full spread. A split between two or three categories usually means different customer groups see the feature differently. A feature where indifferent wins outright can probably leave this year's plan, however loud the request channel is.
A sensible order: close must-be gaps first, because they cost you customers now. Then improve the one-dimensional attributes where you trail competitors. Then pick one or two attractive features that fit your positioning.
Satisfaction and dissatisfaction coefficients
Berger et al. also proposed two numbers per feature. The satisfaction coefficient is (A + O) divided by (A + O + M + I): the share whose satisfaction rises when the feature is there. The dissatisfaction coefficient is (O + M) divided by (A + O + M + I), usually shown with a minus sign: the share whose satisfaction drops when it is missing. R and Q answers stay out of both.
Plot each feature with one coefficient per axis and you see the whole list at once. High on both behaves one-dimensional. High on dissatisfaction only points to must-be, high on satisfaction only to attractive, low on both to indifferent. But a coefficient built on forty answers jumps around. See how many respondents you need for why.
Segments disagree, and that is the finding
A Kano category describes how one group of people relates to a feature. Single sign-on is indifferent to a freelancer and a must-be for an IT department. Automatic translation is attractive to a team surveying in one country and a must-be for a team surveying in six.
Break the results down by the segments your decisions depend on, such as plan, company size or role. A messy overall spread often turns out to be two clean ones stacked together. Ask the segment questions in the same survey, and size the sample so each segment has enough people.
Attractive features do not stay attractive
Categories drift. What delights today is expected tomorrow, because competitors copy it and people get used to it. Kano described this life cycle himself in later work: attractive quality becomes one-dimensional and eventually must-be. Autosave in a form builder was once a nice surprise. Now nobody would accept a builder without it.
So a Kano result has a shelf life. Repeat it about once a year with the same wording. A feature sliding from attractive to must-be can no longer stay half-finished.
One question for the whole product
For the product as a whole, one question has become the standard: "How would you feel if you could no longer use the product?", answered with very disappointed, somewhat disappointed or not disappointed.
It comes from Sean Ellis, who ran growth in the early days of Dropbox, LogMeIn and Eventbrite. Benchmarking nearly 100 startups, he found that companies struggling to grow almost always had fewer than 40 percent answering very disappointed, and companies with strong traction almost always more. In 2018 Rahul Vohra described how Superhuman used it. In summer 2017 their score was 22 percent. Recalculated for only the kind of user who most often answered very disappointed, it was 33 percent, and within three quarters of targeted work the score reached 58 percent.
Two cautions. The 40 percent comes from one person's set of startups, so use it as a reference point and do not treat it as a pass mark. And ask only recent users. Superhuman added open questions about who benefits most, the main benefit, and what to improve. The reasons are in those answers.
Running a Kano study in SurveyLane
Give each feature its own page with a short description or image, then the two questions as single-choice items with the same five options. Use conditional logic to show only features that match a respondent's plan or role; the post on conditional logic and feedback shows how. Randomise the page order so the last features do not always get the tired answers.
End with one optional open question: "Is there anything you would miss that we did not ask about?" Must-be gaps you never thought of show up there. AI-assisted analysis can group them into themes, but check those against the raw text, as the post on analysing open answers explains.
When Kano is the wrong tool
Kano tells you what kind of feature something is. It does not rank features within a category, and it says nothing about price. For a forced ranking of fifteen candidates, MaxDiff does a better job, as the post on measuring what people find important explains. For price, see measuring willingness to pay. And for a concept too new to picture, five interviews with a prototype beat any survey.
Frequently asked questions
How many respondents does a Kano survey need?
There is no fixed number, but category shares behave like any other percentage, so the margin of error depends on group size. With around 100 respondents per segment you can usually see which category dominates a feature, while gaps of a few percentage points between categories need a much larger sample.
Can I use a Kano survey for features that do not exist yet?
Yes, that is the main use, as long as respondents can picture the feature. Describe it as a concrete benefit in one or two sentences, ideally with a sketch or screenshot, and leave out internal names. For entirely new concepts that need explaining, interviews with a prototype tell you more than any survey question.
What do I do with features that come out as indifferent?
Do not drop them automatically. First check whether the result holds in every segment, because a feature most customers shrug at can be a must-be for a small and valuable group. If it is indifferent everywhere, it is a strong candidate to postpone, however often people request it.
Is the very disappointed question a replacement for Net Promoter Score?
No. The very disappointed question asks how much people depend on your product today, and NPS asks whether they would recommend it. The first says more about whether the product has found its core users, the second more about word of mouth, and both depend heavily on who you ask.