Kano, part 3: Making data-driven decisions

Going forward, I plan to post new Product Management topics to Medium instead of my original blog. This is my first post original to Medium.

In my previous posts about Kano, I discuss what it is and why it is important. However, to effectively use Kano without adding your own personal bias, consider crafting a survey so you can map features to each quadrant in the Kano model.

If you’re just starting out, you probably want to spend a good amount of time listening and talking with potential customers. A survey right-out-of-gate will be confusing. Basically, with surveys, use them once you have enough data and are now trying to prioritize and validate the feedback you’ve already collected.

There are many articles out there that describe how to do this (here’s an example). I’ll attempt to summarize the key points and next steps below.

Some prep work:

1. Figure out who to target. If you’re past the MVP stage and actually selling your product, you probably have one — maybe two — key user and buyer personas. In the B2B world, I’d focus on the user for now. Buyers are generally C-level folks who want outcomes, whereas users need features to reach those outcomes.
2. Select the key features (say 5–10) that you’re looking to better understand. I suggest not going past 10 because users, no matter how much they love your product, likely won’t have more than 5 minutes to answer your survey. If you need help figuring out what features to select, check out my post: Doing one thing really well.

Implementing the survey
OK, now it’s time to implement a survey so you can build a data-driven Kano model. For each feature, you’ll need to ask at least 2 questions:

1. First, write a functional question. This is so you can measure the user’s reaction to the existence of this potential feature. An example: At Veracode (my current company), we help developers test their code before shipping. As a result, ‘speed of test’ is an important non-functional metric. Consider this question:
   “If a security assessment takes under 15 minutes to complete, how do you feel?”
2. Second, write a dysfunctional version of the first question. You want to measure their reaction to a scenario where the feature does not exist. Expanding upon the above example:
   “If some security assessments take longer than 15 minutes, how do you feel?”
3. Now, for each question, provide these answers:

• (Like it) I enjoy it that way
• (Expect it) It is a basic necessity or I expect it that way
• (Don’t care) I don’t care
• (Tolerate it) I dislike it, but I can live with it that way
• (Dislike it) I dislike it, and I can’t accept it

Measuring results
After you’ve run the survey, you need to analyze results. For each ‘feature’, measure responses using the two functional and dysfunctional questions.

For example, let’s assume I received the following answers to the ‘speed of test’ question:

• Functional version — Under 15 mins: Like it
• Dysfunctional version — Over 15 mins: Tolerate it

Using the above table, we will see that our response maps to ‘A’. Since you’ve probably surveyed more than 1 person, tally all your responses like this:

• Q — 0
• A — 10
• P — 7
• R — 2
• I — 1
• M — 0

Total people surveyed: 20

If we apply a simple max() function to this set of results, we determine that ‘A’ has the most votes.

Now what do these letters even mean?

• A — Attractive: Customers are delighted by this, and may not have expected to see this in your product. (You’re going to impress some people with this.)
• Q — Questionable: There is probably something wrong with how you worded this question. How can someone like both the positive and negative forms?
• I — Indifferent: Users don’t care if you have it or not. (In my experience, really study why users don’t care — this is where qualitative interviews help.)
• M — Must be: Basic expectation features. Table stakes. (If you don’t have this, you probably won’t be invited to the evaluation.)
• P — Performance: Customers like it when it’s there, and dislike it when it’s not. The more you invest here, the happier the customer. (Prospects generally evaluate these things during the sales process.)

Conclusion
We can conclude that ‘speed of test’ is an attractive quality. In other words, it’s a delighter if your security tests run in under 15 minutes!

• A — Attractive: Maps to the upper-left quadrant (Delighters)
• P — Performance: Maps to the upper-right quadrant (Performance needs)
• M — Must be: Maps to the lower-right quadrant (Basic needs, basic expectations)
• I — Indifferent: Maps to the middle of the graph (where X and Y axes meet).

Going forward: Making decisions with this data
You will probably find it important to close ‘must have’ and ‘performance’ gaps than features that trigger an ‘indifferent’ reaction. ‘Indifferent’ or ‘questionable’ features require additional inspection. Maybe they are important to a specific segment of your customer base, but not all. Depending on the $ value of that segment, you may or may not want to invest in those features.

So what about ‘attractive’ qualities? Attractive qualities are important, but if you are missing key features to solve the customer job, then close those gaps first.*

*Note: This is not an absolute rule. When the Apple Watch first launched, there was no way to observe the time if the watch was simply sitting on your desk charging. One may consider this a ‘must have feature’ relative to analog watches. So, if you’re pioneering a new space or disrupting the status quo, different rules will apply.

And… some additional tips on why/when you should use this model

• You are building a new product, and want to understand what users want
• You are in a very competitive space, and you think you may have some ideas that differentiate you. Use this to see if they are really attractive features.
• You have launched a few features and they haven’t been big hits. Try and figure out what customers really want using this model.
• You don’t need surveys to do this; you can also leverage existing A/B tests in your product. In fact, you’ll get higher quality feedback if you conduct tests with a real product.

There are many versions and interpretations of the Kano model on the web. A great summary of this can be found at FoldingBurritos.com, which inspired me to write this summary: https://foldingburritos.com/kano-model/