Kano analysis techniques made simple

You've run a Kano survey and you've got all the responses from your participants. But now what - how should you analyse them?

Start here for a primer on how the Kano model technique works, then continue below for more detail on Kano analysis - how the different analysis approaches actually work, how to apply them, why they're different, pros and cons etc.

Method #1: Categorisation analysis (simple)
Method #2: Continuous scale (advanced)

This article is quite technical, but hopefully simple enough that you can understand how you can apply the different techniques to interpret what the results tell you about customer satisfaction.

Method #1: Categorisation analysis (simple)

This approach is the simplest to understand and apply.

In a nutshell, every response to a feature is categorised and then we look for the most common category across all the responses (ie. the modal average). That becomes the primary category for that feature.

Here's an example.

Feature: "heated seats in a car"

Responses:

#	Positive	Negative	Category
1	I like it	I expect it	Delighter
2	I am neutral	I am neutral	Indifferent
3	I dislike it	I can tolerate it	Reverse
4	I like it	I am neutral	Delighter
5	I dislike it	I like it	Reverse
6	I like it	I can tolerate it	Delighter

Each response is like a vote for a particular category. And if we count up these different votes...

Category	Count
Delighter	3
Indifferent	1
Reverse	2

Delighter has the most votes, so the feature is categorised as a Delighter.

This type of Kano analysis is simple, easy to understand, easy to compute by hand (or with a Kano analysis template if you prefer), and gives nice clear answers.

However it has a couple of drawbacks.

The wnner could be marginal - In our example above, 50% of the votes were for Delighter. This seems like a reasonable number. But, Delighter only got 1 more vote than Reverse overall so it's not a clear winner by a long distance.
Other strong opinions are obscured - In the example there are 2 people that feel very strongly against the feature, but their opinions would be invisible after the categorisation has taken place.
Final categories are a narrow view - All the information that's represented by the other votes doesn't contribute to the overall category because it's just based on a simple count. You lose any possible nuance.

KanoSurveys.com tries to solve the first two problems through two innovations:

Statistically significant winners - We don't just count the votes in the responses, we look for a winner that is clearly in front of the pack, by using empirical rules of normal distribution. This avoids marginal winners.
Radar charts - We show the distribution of responses between the different categories in a visual way. This is quicker to understand at a glance than a table of numbers, and can tell you very quickly if there are any other strong responses in the data.

We solve the third problem by offering a second type of Kano analysis - a continuous scale which utilises all the different responses to produce a more representative view of customer satisfaction.

Method #2: Continuous scale (advanced)

Whereas the first method is like a modal average, this is more like a mean average (this is the type average you're probably most familiar with).

This approach, developed by researcher Bill DuMouchel, uses all the responses to contribute to the final category so all the nuance is included, and the difference in strength of expression between similar responses can be brought out.

Each response is first converted onto a numerical scale, with more weight attached to positive end of the scale.

	Dislike	Tolerate	Neutral	Expect	Like
Feature is present	-2	-1	0	2	4
Feature is absent	4	2	0	-1	-2

These scales are designed to correlate with potential satisfaction if the feature were included. So "Disliking" the absence of the feature suggests the participant would be satisfied if it did exist.

Next, all functional and dysfunctional responses are averaged together to give a single pair. This pair represents the aggregate total of all the different opinons of the participants.

The pairs for all features are plotted onto a grid, with the different areas of the grid mapped out to Kano categories.

It's possible to use the grid definitions to just declare a single category for each feature but again that would obscure some of the nuance that is plain to see when you plot onto the grid.

To return to our previous example, the scores would look like this:

#	Positive	Negative
1	I like it = 4	I expect it = -1
2	I am neutral = 0	I am neutral = 0
3	I dislike it = -2	I can tolerate it = 2
4	I like it = 4	I am neutral = 0
5	I dislike it = -2	I like it = -2
6	I like it = 4	I can tolerate it = 2
Average	1.33	0.17

And if we pop that onto the grid, we get:

Screenshot showing an example continuous maxtrix analysis

In this example you can see that due to the strength of the reverse responses the overall average was evenly balanced between Delighter and Reverse, leading to an Indifferent.

In some ways this is more representative of the total audience opinion.

There are definitely some people who would be delighted, but there are also some who would be upset.

In this way the continuous analysis makes sure you're not ignoring anyone who didn't vote with the crowd.

Method #3: Satisfaction coefficients (advanced)

The last type of analysis is a way of getting an overall estimate of how customer satisfaction would change on an absolute scale, if the feature were present or not.

This is helpful if you have many features within a single category, to rank them in order of either minimising negative impact of not having the feature, or maximising positive impact of having it.

The coefficients are calculated by adding up the counts of the main Kano categories (delighter, performance, must have and indifferent) and using that as the base.

Then we look at what proportion of that base is likely to be either happy or sad:

For the "better" coefficient (to assess customer satisfaction potential if the feature were present), we see what proportion of that base voted for either delighters or performance. These are the categories where people will feel some level of happiness that the feature is present and well executed.
For the "worse" coefficient (to assess customer satisfaction impact if the feature were absent), we look at the proportion of people who voted for either performance of must have. Anyone in these categories will be disappointed to some extent if the feature wasn't there or is not good quality.

So, for each feature we would do this:

total = (delighters + performance + must_have + indifferent)

better = (delighters + performance) / total

worse = ((performance + must_have) / total) * -1

The "worse" coefficient is always expressed as a negative number. When you draw it on a graph it looks like this:

Screenshot showing a satisfaction coefficient graph

Summary

Well, hopefully that wasn't too technical and now you have a better understanding of the pros and cons of each type of Kano analysis.

To sum it up - the simple categorisation will give you the majority view. We use statistical significance methods to make sure you're seeing a real winner.

The continuous analysis gives you the aggregate view taking into account all the different opinions in the responses, and the satisfaction coefficients give an indication of how satisfaction could change on a linear scale.

Which is better?

Well, we think they're all useful in different ways and worth considering together to give a holistic view of customer satisfaction.