Kano Model Survey Guide — Telling Delighters from Must-Haves

"Let's build the most-requested features first." It sounds obviously right, and in product development it misfires constantly. The reason: a feature that makes satisfaction spike and a feature that's taken for granted until it breaks (and then people are furious) run on completely different logic. Polish the first one all you like and complaints won't disappear; polish the second and satisfaction won't budge.

Satisfaction is asymmetric. The Kano model captures that asymmetry with five quality types and changes how you act on each feature accordingly. Proposed by Japan's Noriaki Kano in the 1980s, it remains in everyday use for product development and service design worldwide. This guide walks through what the five categories mean, the Kano model's distinctive "two questions per feature" format, classification with the evaluation table, visualization via the Better-Worse coefficient, and how it connects to IPA and key driver analysis — all with a practitioner's feel for what actually matters.

1. Why "satisfaction asymmetry" matters

Intuitively, we assume "better quality means higher satisfaction." Reality doesn't work that way. Kano et al. (1984) showed that for some quality attributes, the relationship between how well something is delivered and how satisfied customers feel is not a straight line.

Take an example. A smartphone whose "battery lasts all day" is just expected. It doesn't impress anyone when it lasts, but it generates intense dissatisfaction when it doesn't. Meanwhile, an "unexpected convenience feature" troubles no one by its absence, yet when it's there, satisfaction spikes — "oh, nice." Both are 'features,' but their effect on satisfaction is exactly opposite.

Ignore this asymmetry and decide your development priorities on "number of requests" or "average satisfaction" alone, and you'll either pour effort into must-be quality and watch satisfaction stay flat, or miss an attractive feature and let a competitor pull ahead. The Kano model is the tool for telling apart how each feature acts on satisfaction.

Importance-Performance Analysis (IPA) and key driver analysis both touch on "satisfaction asymmetry (Matzler 2004)," but what makes the Kano model distinctive is that it goes after that asymmetry directly at the question-design stage.

2. Kano's five categories — sorting features by how they act on satisfaction

The Kano model classifies quality attributes into five types.

What each category tells you to do

• Must-be quality: meet it reliably as the "floor." Competing here won't raise satisfaction (defense)
• One-dimensional quality: satisfaction grows with the level of investment. The battleground against competitors (grow)
• Attractive quality: even a couple of these, if they land, become differentiation. The next headline feature (offense)
• Indifferent / reverse quality: decide not to do it, or to cut it. Save resources and strip out over-engineered features

The core of Kano is that "many requests = should build it" is wrong — the meaning of the investment changes with the type.

3. The Kano model's distinctive "two questions per feature" format

What sets the Kano model decisively apart from other surveys is that it asks two paired questions about each feature. These are the functional question and the dysfunctional question.

The question form

For a given feature, you ask the following two questions.

[Functional question] How do you feel if this feature IS present?
[Dysfunctional question] How do you feel if this feature is NOT present?

Both questions use the same shared five-point answer scale.

1. I like it (satisfied)
2. It must be that way (it should be expected)
3. I'm neutral (no particular feeling)
4. I can live with it (tolerable)
5. I dislike it (dissatisfied)

The feature's type is determined from the combination of answers to the "present" and "not present" cases. "Present → I like it / Not present → I'm neutral" means attractive quality; "Present → it must be that way / Not present → I dislike it" means must-be quality, and so on.

Why two questions are needed

Ask satisfaction with a single question — "Is this feature important?" — and the asymmetry stays invisible. Even when someone answers "important," you can't tell whether that's "I'm in trouble without it (must-be)" or "I'm glad to have it (attractive)." Only by closing in from both the functional and dysfunctional sides does the type of effect on satisfaction become clear. This is the heart of Kano's question design.

How you phrase the questions sways the results. For the principles of avoiding leading wording and double negatives, see the complete guide to writing survey questions.

4. Classifying with the evaluation table

The combinations of the functional question (5 options) × dysfunctional question (5 options) come to 25. You map these onto the Kano evaluation table to classify each respondent and each feature into a type.

How to read the representative cells:

• The basics of aggregation: for each feature, take "the most frequently assigned type (the mode)" as that feature's type
• Questionable (Q) answers: logically contradictory combinations such as "Present → I like it / Not present → I like it." Candidates for removal during data cleaning. If there are too many, respondents may not understand the question (data cleaning guide)

The standard aggregation is to produce a distribution of types per feature (what % is A, what % is M, and so on) and represent the feature by its modal type.

5. Visualizing with the Better-Worse coefficient

Modal classification is simple, but it leaves you stuck on features that split by a hair, like "attractive 45% / one-dimensional 40%." That's where the Better-Worse coefficient (CS coefficient / Customer Satisfaction coefficient) proposed by Berger et al. (1993) comes in, visualizing the result as a continuous quantity.

The formulas

• Better coefficient (satisfaction coefficient) = (A + O) / (A + O + M + I)
• Worse coefficient (dissatisfaction coefficient) = −(O + M) / (A + O + M + I)

(A = attractive, O = one-dimensional, M = must-be, I = indifferent — the count of each answer)

• Better is "how much satisfaction rises when you provide the feature" (0 to 1; larger means a bigger satisfaction-lifting effect)
• Worse is "how much dissatisfaction grows when the feature is missing" (−1 to 0; the larger the absolute value, the bigger the hit from its absence)

Into four quadrants on a scatter plot

Plot Better on the horizontal axis and Worse (absolute value) on the vertical axis, and the features separate into four zones.

• High Better, high Worse → one-dimensional quality (the main battleground where more investment pays off)
• High Better, low Worse → attractive quality (a candidate headline differentiator)
• Low Better, high Worse → must-be quality (the foundation to protect)
• Low Better, low Worse → indifferent quality (deprioritize)

This presentation is conceptually close to IPA's four quadrants, letting an executive meeting debate "how much to invest in which feature" from a single chart. For drawing the scatter plot, see the survey data visualization guide.

6. When to use Kano vs. IPA vs. key driver analysis

The Kano model is complementary to other satisfaction-analysis methods. Don't conflate them — choose by purpose.

• Kano model: classifies each feature's type of effect on satisfaction (attractive / must-be / one-dimensional) directly at the question-design stage. Good for selecting new features and designing roadmaps
• Key driver analysis (KDA): derives, from existing satisfaction data, the factors that statistically move overall satisfaction. Regression-based. Good for identifying improvement drivers for a service already in operation
• Importance-Performance Analysis (IPA): maps improvement priority across four quadrants of importance × satisfaction. Good for taking stock of the current state

Standard combinations

• Pre-launch / evaluating new features → use Kano to size up feature types, meet must-be quality while seeding in one or two attractive features
• Improving a live service → use KDA to identify the drivers that matter, and IPA to map priority. Use Kano's "must-be / attractive" lens to spot "must-be quality that won't raise satisfaction even if you fix it"

Where KDA / IPA "analyze the satisfaction structure of what already exists," Kano is the pre-launch, planning-oriented tool for "sizing up the type of a feature you're about to build or ship" — that's the division of labor. For pre-launch feature prioritization, it's also used alongside MaxDiff and conjoint analysis.

7. The editor's view — five things you must not do with the Kano model

From the vantage point of continuously following industry cases and the voices of practitioners, here are five accidents that recur with the Kano model.

1. Expecting satisfaction to rise by "strengthening" must-be quality

The most common misconception. Must-be quality is the type that's expected when met, resented when missing. Polish it harder than anyone and satisfaction still won't rise. "Push login reliability to the absolute limit and customers will be moved" doesn't happen. Must-be quality is the "don't let it break" defense, not something to grow. To go on offense, head for one-dimensional and attractive quality.

2. Assuming every "most-requested feature" is one-dimensional quality

Most features customers say they "want" are one-dimensional, but some are attractive quality (hard to put into words) and some are must-be quality (so obvious it never shows up as a request). Decide development on the request-count ranking alone and you'll miss must-be gaps and overlook attractive quality. Judge by Kano type, not request volume.

3. Leaving contradictory (Questionable) answers in place

Mixing logically contradictory answers like "Present → I like it / Not present → I like it" into the aggregation without removing them. A feature with many Q answers is a sign the question's explanation isn't getting through. You need to revise the feature's description or take the call to drop it from the analysis. Skip the cleaning and your classification wobbles.

4. Overconfidence in attractive quality as "permanent differentiation"

Today's attractive quality becomes tomorrow's must-be quality (Kano himself flagged this as "quality obsolescence"). A smartphone camera was attractive quality when it debuted; now it's must-be quality. Kano classification is a point-in-time snapshot, and unless you re-measure periodically you'll miss the moment a former headline feature turns into "of course it's there."

5. Ignoring segments and settling on one classification overall

A feature can be attractive quality for heavy users yet indifferent quality for light users — this happens routinely. Round it down to one type with an overall aggregate and the differences between segments vanish. For important features, combine with customer segmentation and look at the Kano classification by segment.

8. Running a Kano model survey with the survey tool Kicue

A Kano model survey splits into a "design the two paired questions and collect responses" phase and an "evaluation-table classification and Better-Worse coefficient calculation" analysis phase. What Kicue handles is the former.

• Two-question (functional / dysfunctional) design: supports a layout that lines up the functional and dysfunctional questions for each feature on a shared five-point scale (question types). You can build one pair per feature
• Presenting feature descriptions: you can design the flow to present each feature's description as text before asking the two questions
• CSV export with respondent IDs: outputs functional and dysfunctional answers one row per response, in a structure you can feed straight into evaluation-table classification and coefficient calculation
• Respondent screening: opening screeners to narrow to your target customers (screening question design guide)

⚠️ What Kicue cannot do

• No automatic Kano evaluation-table classification: the 25-cell matrix classification is done after CSV export, in Excel / R (e.g., packages such as KanoR) / Python. Kicue itself doesn't carry a Kano classification feature
• No Better-Worse coefficient calculation or scatter plot either: coefficient calculation and visualization are done in Excel / R / Python
• No automatic Questionable (Q) answer detection either: cleaning and removal are post-export processing
• No segment-level Kano classification either: per-segment aggregation is done in an external tool

As related reading, pairing the Importance-Performance Analysis (IPA) guide, the key driver analysis guide, the MaxDiff design guide, conjoint analysis in practice, and the concept test survey guide brings into view the whole picture of pre-launch, product-development research: "size up the type of a feature (Kano) → measure priority (MaxDiff / conjoint) → evaluate the concept."

Summary — six points for mastering the Kano model

1. Satisfaction is asymmetric — polishing must-be quality won't raise satisfaction. Go on offense with one-dimensional and attractive quality
2. Ask two questions per feature — determine the type from the combination of functional and dysfunctional questions. A single question can't reveal the asymmetry
3. Decide by type, not request count — many requests don't mean rising satisfaction if it's must-be quality
4. Visualize with the Better-Worse coefficient — turn features that split on the mode into a continuous quantity and make investment decisions across four quadrants
5. Kano classification is a point-in-time snapshot — attractive quality eventually becomes must-be. Re-measure periodically
6. Look at it by segment — don't round down to one classification overall; combine important features with segmentation

The Kano model isn't an "elaborate analytical technique" — it's a tool that takes one insight, "satisfaction is asymmetric," and bakes it into question design. Get just that distinctive two-questions-per-feature format designed correctly, and you can lift the quality of product-development decisions a notch — from "build in order of most requests" to "size up the type and invest accordingly."

If you want to design a Kano model survey for a product or feature, why not try the free survey tool Kicue? From the two-question functional/dysfunctional design, to presenting feature descriptions, to CSV export with respondent IDs, you can start the input-data part of a Kano survey on a single account (evaluation-table classification, Better-Worse coefficient calculation, and scatter plots are run in combination with Excel / R / Python).