Kano Model: A Worked Survey, Classification Matrix, and Product Decision

Start with one feature and one decision

Consider a walking-route app deciding whether people should be able to download a saved route before leaving home and reopen its map without a connection. This worked example follows that decision from survey wording to release scope. The response counts are illustrative inputs you can use to practice the analysis.

The decision is narrower than “Should we build offline mode?” That phrase could mean a saved map, live navigation, route search, or emergency services. A respondent who imagines one of these is answering a different question from someone who imagines another. Describe the behavior, its boundary, and what remains available without it.

Microsoft Research’s Kano guide demonstrates the paired-question approach with a small food survey and recommends asking why people chose their responses. Its useful lesson is that wanting an idea and being dissatisfied without it are different signals. Our route example adds the operational steps: coding the pairs, separating segments, and making a bounded product decision.

Ask both questions without selling the feature

Show the same short feature description to everyone. Ask: “If you could download a saved route and its map beforehand and reopen them without a connection, how would you feel?” Then ask: “If you could only reopen that saved route and its map while connected, how would you feel?” The second question describes absence; it does not add an unrelated penalty.

For each question use the same five choices: I like it; I expect it; I am neutral; I can tolerate it; I dislike it. Do not substitute a simple importance score. The pair, rather than either answer alone, determines the category.

Pilot the wording before recruitment. Ask a pilot participant to explain what downloading would and would not include. If they assume turn-by-turn navigation, repair the description before collecting the main responses. Add “What situation were you thinking of?” after the pair, and record whether the person usually walks in rural areas or cities using a defined, consistent recruitment question.

Classify the answer pair with a matrix

Read a row for the answer when the feature is present, then a column for the answer when it is absent. A means attractive, O means one-dimensional or performance, M means must-be, I means indifferent, R means reverse, and Q means questionable. Keep the raw answers alongside the code so another researcher can check your work.

Like-present and dislike-absent produces O: the respondent values having the feature and misses it without it. Expect-present and dislike-absent produces M: its absence violates an expectation. Like-present and neutral-absent produces A: it is welcome but not essential to that person.

Q is not a product category to average away. Review these pairs for misunderstood wording, inconsistent answers, or an overly broad feature description. R can reveal a real preference against the behavior; interview those respondents instead of automatically throwing them out.

Inspect the distribution before choosing a label

In this practice dataset, 24 people answer both questions. Twelve usually follow rural routes and twelve usually follow city routes. The table shows all classification counts, including the questionable pair. Each row must sum to the number of respondents in that segment.

Across everyone, M has 9 responses, A has 7, O has 4, I has 3, Q has 1, and R has 0. Calling offline access simply “a must-have” would conceal the city group’s different pattern. Within the rural group, 8 of 12 classify it as M; within the city group, A is the largest category with 6 of 12. These small counts guide a follow-up conversation; they do not establish the preferences of all walkers.

The single Q belongs to the city group. Keep it visible and investigate it. If you show percentages among valid classifications, label the denominator: there are 23 non-Q responses overall and 11 in the city group. Do not quietly use 24 for one percentage and 23 for another.

Turn the finding into a release decision

For a release explicitly aimed at rural walkers, the team would investigate offline reliability before spending the same effort on cosmetic map themes. The reasoning is specific: this segment expects a route to remain available after connectivity disappears. The next research question is when and how they prepare for that situation.

Prototype download selection, progress, completion, and reopening without connectivity. Separately investigate storage size and map-data licensing. Kano answers cannot prove that a download implementation works, that a participant will remember to use it, or that licensing permits distribution. These uncertainties need different evidence.

Write the first version as a complete outcome: a person selects one saved route while connected, sees the download complete, switches to airplane mode, and reopens the same map. Show a clear unavailable state for routes not downloaded. Leave offline search and live navigation outside this release rather than letting “offline mode” absorb them.

Copy this worksheet for your own feature

Run one clearly described behavior through the worksheet at a time. Save the questionnaire version with the results; otherwise a later survey can appear to show changing expectations when the wording actually changed. Repeat studies should explain changes in audience, recruitment, and feature description.

Finish with a decision and a next check. “Attractive” is not an instruction to implement immediately, and “indifferent” is not permission to remove a dependency required by another feature. Use the classifications to understand customer expectations, then judge the release against its purpose and constraints.

References

• Microsoft Research: a practical Kano survey
• Qualtrics: Kano analysis