Enter how many respondents gave each answer for a feature, or paste the raw pairs from a survey, and this classifies the feature, works out its Better and Worse coefficients, and plots it against every other feature on the list. Nothing is sent anywhere.
| Feature | A | O | M | Ind | R | Q | Category | Better | Worse | Remove |
|---|
A, O, M, Ind, R and Q are respondent counts: how many people's answers landed in Attractive, One dimensional, Must be, Indifferent, Reverse and Questionable once each person's functional and dysfunctional answer is looked up in the evaluation table below. Type counts straight into those columns, or use the paste box to classify raw pairs for you.
Each line is one respondent's two answers, separated by a comma: like, expect, neutral, tolerate or dislike (or just the first letter). Lines that do not parse are skipped.
How to write the question pair
Every feature needs two questions, asked together, never one without the other.
The functional question asks how the respondent would feel if the feature were there. The dysfunctional question asks how they would feel if it were not. Both use the same five answers: like it, expect it, neutral, can tolerate it, dislike it.
Worked example, for a feature called dark mode on a feedback board:
If this tool had dark mode, how would you feel? If this tool did not have dark mode, how would you feel?
Each respondent picks one of the five answers for each question. The pair of answers, not either answer alone, is what the evaluation table above turns into Attractive, One dimensional, Must be, Indifferent, Reverse or Questionable. A respondent who says they would like dark mode and could also tolerate its absence lands on Attractive. A respondent who expects it and would dislike its absence lands on Must be.
How many respondents it takes
A commonly cited floor is about 15 respondents in each customer segment you plan to read on its own, since a segment is really its own smaller survey. Fewer than that and one person changing their answer can flip the category. The Category column above shows the respondent count next to the category name so a thin result is visible, not hidden behind a confident sounding label.
How to read the plot
Better runs up the page, from 0 to 1: how much satisfaction rises when the feature is present. Worse runs across, from 0 to 1 as an absolute value: how much dissatisfaction rises when it is absent. A feature in the top right pulls on both and behaves like a straight line, more is better. Top left is Attractive: nobody is upset by its absence, but its presence earns credit. Bottom right is Must be: its absence is punished and its presence earns nothing extra. Bottom left is Indifferent: the feature barely moves either number.
Two mistakes
Averaging the categories. A, O, M, Ind, R and Q are not points on a scale, so there is no such thing as the average of three respondents who said Attractive and two who said Must be. The feature's category is whichever letter the most respondents landed on, which is what this tool shows. The Better and Worse coefficients come from the counts themselves, never from an averaged category.
Ignoring the Questionable answers. A respondent lands on Questionable when their two answers contradict each other, such as liking the feature present and also liking it absent. A feature with a lot of Questionable answers usually means the question was unclear or respondents misread it, not that the feature is safely Indifferent. Check the Q count before trusting a low Better and low Worse score.
Where the features to ask about come from
Picking which fifteen or twenty features are worth a Kano survey is a separate problem from running the survey, and a request board is one way to shortlist them. VoteFirst is a feature request board where visitors post and vote with no account required. It does not run Kano surveys or ask the dysfunctional question; it only counts who asked for what. The requests with the most votes, or the most votes from paying customers, are a reasonable shortlist to put in front of a Kano survey. Fourteen days on Lite or Pro, a card at signup, then $9 or $39 a month.
Is there a questionnaire and analysis template?
Yes. Take the xlsx. One sheet lists the functional and dysfunctional question for each feature with the five answers as a dropdown, a second sheet tallies the 25 answer pairs per feature with the same formulas as the tool above, already filled in for three example features, and a third sheet explains how to use it. It opens in Excel, Numbers, LibreOffice and Google Sheets.