Kano, worked
Run a 40-respondent Kano survey through the classification table, read the three verdicts it returns, and say what an attractive feature is worth before and after a competitor ships it.
On this page
A Kano survey answers a question RICE cannot: not how much a feature is worth, but what kind of worth it has. Forty respondents and two questions per feature classify each one as something customers assume, something they would pay more for, or something they do not care about either way — and those three are built, sold and declined in three different ways. The arithmetic is a tally, which is why it holds up in a roadmap meeting where a score would be argued with.
Two questions per feature
Kano, Seraku, Takahashi and Tsuji ("Attractive Quality and Must-Be Quality", 1984) ask each respondent the same pair of questions about each feature. The functional form: "If the product exported ticket data to CSV, how would you feel?" The dysfunctional form: "If the product did not export to CSV, how would you feel?" Each takes one of five answers: I like it, I expect it, I am neutral, I can tolerate it, I dislike it.
The pair of answers places the respondent in one cell of a five-by-five table, and the cell has a category:
- Must-be (M) — expect, neutral or tolerate if present; dislike if absent. Its presence earns nothing; its absence disqualifies.
- One-dimensional (O) — like if present, dislike if absent. More of it is better, in proportion; this is what performance features look like.
- Attractive (A) — like if present; expect, neutral or tolerate if absent. Nobody misses it, everybody likes it: the differentiator.
- Indifferent (I) — neither answer carries feeling.
- Reverse (R) — the respondent prefers not having it. Worth a segment check, since some people want the opposite.
- Questionable (Q) — like both having and not having, or dislike both; the question was misread and the response is dropped.
A feature's category is the one most respondents land in. The survey below went to 40 people across the support tool's tenants — 25 agents and 15 support leads — on three candidates from the backlog.
Forty answers, tallied
CSV export. Two respondents answered like to both forms and are dropped as questionable, leaving 38. Of those, 27 said expect-or-neutral if present and dislike if absent: must-be. Six were one-dimensional, two attractive, three indifferent.
export: M 27 O 6 A 2 I 3 (n = 38, 2 dropped) 27 ÷ 38 = 71% must-be
Auto-summarise. Three dropped, 37 valid: 23 attractive, 5 one-dimensional, 1 must-be, 8 indifferent.
summaries: A 23 O 5 M 1 I 8 (n = 37, 3 dropped) 23 ÷ 37 = 62% attractive
Dark mode. All 40 valid: 22 indifferent, 11 attractive, 4 must-be, 3 one-dimensional.
dark mode: I 22 A 11 M 4 O 3 (n = 40) 22 ÷ 40 = 55% indifferent
Two coefficients from Berger and colleagues (1993) compress each row into a pair: satisfaction if present, (A + O) ÷ (A + O + M + I), and dissatisfaction if absent, (O + M) ÷ (A + O + M + I).
export: satisfaction 8 ÷ 38 = 0.21 dissatisfaction 33 ÷ 38 = 0.87
summaries: satisfaction 28 ÷ 37 = 0.76 dissatisfaction 6 ÷ 37 = 0.16
dark mode: satisfaction 14 ÷ 40 = 0.35 dissatisfaction 7 ÷ 40 = 0.18
Read as: shipping export pleases almost nobody and its absence angers 87% of the valid respondents (33 of 38); shipping summaries pleases three in four and its absence bothers one in six; dark mode does little either way.
What each category tells you to do
The must-be is built because not having it costs renewals, not because it earns them. That is why the support lead who runs 1,200 agents asked for export three quarters running (the RICE lesson scored it at 6,000 and it won there too); nobody will cite it in a case study, and the 0.87 says a tenant evaluating alternatives will strike the product off the list without it. A must-be is never a launch headline, and never the thing an add-on is priced on.
The attractive is what an add-on is priced on. Summaries at 0.76 satisfaction and 0.16 dissatisfaction is the shape of a feature customers would pay extra for and would not leave over — priced like the pricing lesson's drafting add-on at $5 a seat, against a token bill of its own. An attractive feature is sold; a must-be is included.
The indifferent is declined, politely and with the number. "55% of the 40 people we asked had no feeling either way, and 11 liked it" ends the dark-mode conversation faster than any score, because the people who wanted it were in the room when the survey went out. If it is cheap enough to build in the gaps, build it in the gaps; never give it a quarter.
Two cautions on the numbers. Forty respondents is a classification instrument, not a measurement: 62% and 55% are 23 and 22 people, and the categories are stable while the percentages are not — a lift from 55% to 62% is three people changing their minds. And the mode hides the split: 11 of 40 found dark mode attractive, and if all 11 are night-shift agents, that is a segment with a real need and a small reach, which is a RICE question.
What "attractive" is worth, before and after
Kano's own observation is that categories move. A feature enters as attractive, becomes one-dimensional as customers learn to expect it, and ends as must-be once every alternative has it. The survey is a photograph of one moment in that drift, and the moment matters.
Before any competitor ships summaries, the feature is attractive: it earns the add-on. Assume the 8,000 opted-in agents take it at $5: $40,000 a month of revenue that exists because the feature does. After the two largest competitors ship summaries, the same survey run twelve months later reads one-dimensional at best: the 0.16 dissatisfaction climbs toward export's 0.87, and the feature's job changes from winning the add-on to not losing the base. The price it can carry falls toward the price of the base product — the add-on line erodes — while the token bill that produces it does not fall at all.
So an attractive feature has a window, and the window is the argument for building it before the sheet says it is the cheapest thing to build. That is a sequencing question with a date in it, which is what cost of delay is for. The survey itself is cheap to repeat — forty people, two questions, an afternoon — and the continuous interviewing cadence already puts three to five of them in front of you a week, so the photograph can be retaken each quarter.
In an interview
The prompt is usually "how would you decide what goes on the roadmap", and a Kano answer is tested on whether you can run the survey and read it, not on whether you can name the categories. Describe the two questions, say what the pair of answers classifies, and give the tally: "27 of 38 must-be, so export goes in and never on a slide." Then show the drift: "summaries are attractive now, and attractive features have a window; in a year this is a must-be and the add-on price goes with it." An interviewer hearing that knows you have run one.
The mistake that loses points is treating the mode as a measurement — reporting "62% attractive" as though the next survey could show 65%, or comparing it against another feature's 55% as a lift. The survey classifies; it does not score. The number to carry out of it is the category and the two coefficients, and the decision each category implies.
Check yourself
1. A fourth feature, SSO, returns M 14, O 15, A 4, I 5 from 38 valid responses. What is it, and what do you do with it?
One-dimensional by a single vote (15 against 14), which is a tie in all but name on 38 people: it sits on the border between must-be and one-dimensional, and the dissatisfaction coefficient settles it — (15 + 14) ÷ 38 = 0.76, absence angers three in four. Treat it as a must-be for the enterprise tier: build it, include it, do not price it.
2. Dark mode came back 11 attractive, and all 11 are night-shift agents at two tenants. Does the 55% indifferent verdict stand?
For the product, yes; for those two tenants, no — and the instrument changes. Kano has classified the whole population; the 11 are a segment with a reach of about 400 agents (assume two tenants' night shifts). Score it with RICE at that reach: 400 × 1 × 0.8 ÷ 0.5 = 640 — a small, cheap item with one judged column, which fits in a gap, not a quarter.
3. A competitor has just shipped summaries and sales reports losing two deals on it. Does the Kano category change this week?
The category is a property of the market, and it has started moving; what changes this week is the decision. The feature is still attractive to most respondents but it is now on a clock, so the question is no longer "is it worth building" (yes, 0.76) but "what does each month without it cost" — and that is cost of delay, with the two lost deals as the first data point for the value-per-month input.