The goal of product managers and user experience designers is to determine which functions can make users more satisfied, and use this information to rank product functional requirements. In order to achieve our goal, we need to consider some important details.
This section reflects the experience and lessons gained by many researchers in practicing Carnot model, including each step:
Select functions and suitable users;
Access to user data (as effectively as possible);
Analyze data.
The first thing to consider is the scope of analysis-including functions and users.
The function to be studied should be the function from which users can benefit. There may be different types of functional requirements in the alternative pool, such as technical debt settlement, sales or marketing team reporting system, design update, etc. But these are beyond Kano's analysis.
Kano model studies users' satisfaction with perceivable and operable functions, but the functions of products are far more than that. If you want to get supporting data that can reject some demands of internal stakeholders such as management, it is dangerous to use Kano model, which will confuse your team, users and yourself.
In addition, the number of functions included in the questionnaire should be limited, especially if there are voluntary users. This can improve the user's participation and attention.
When selecting users (or potential users) to participate in the research, we must consider the demographic characteristics, social groups or roles to which users belong. Otherwise, the final data is likely to spread out the whole map.
Users/potential users may have different qualities and different feelings about functions. However, if some groups to which they belong are considered, the noise in the analysis can be significantly reduced.
Jan Moorman realized the importance of this when introducing the functions of new products to potential users. The core function of this product already exists, which is widely known (probably) because of competitors' products. Nevertheless, some users still think it is attractive, while others think it is necessary. Then she concluded that these different reactions were due to users' sensitivity to the market. When she decomposes the user's answer according to the user's personal data, the result of each function becomes clearer.
Although there are various possibilities of segmentation, we must choose the segmentation that is meaningful to the product. Suppose you work in B2B SaaS. Consider adding a function that allows users to associate invoices with purchase orders. This function has very different appeal to small enterprises and large enterprise customers.
Keep this in mind whether you are choosing users to study (because you know the goal of the function) or the data obtained after analysis.
The questionnaire and its display mode are the only input of Kano research. Therefore, ensure the effectiveness of this step.
The key to designing the questionnaire is to ask as clear and concise questions as possible. Each question should represent only one function. If the function is complex and contains multiple steps and sub-processes, it needs to be split.
Problems should be expressed from the perspective of users' interests, not from the perspective of product functions. For example, "What do you think if you can choose to improve your photos automatically?" Than "If you have MagicFix? What do you think? " .
At the same time, pay attention to the polarity wording of the question pair. In other words, the problem of functional deficiency is not necessarily the opposite of functional problem, it is just functional deficiency. Watch an example of this video editing application, which considers optimizing its export speed:
Function question: "What do you think if the time to export any video is within 10 second?"
Error function missing question: "What do you think if it takes more than 10 seconds to export any video?"
A more appropriate problem of lack of function: "What do you think if it takes more than 10 seconds to export some videos?"
It is better to show the user the function as directly as possible and then ask her what she thinks when she has or does not have the function, than to ask a clear question.
You can use words to describe functional advantages, and then use prototypes and interactive wireframes or models instead of words. Through visual and dynamic "explanation", users can understand these functions more clearly and ask questions.
If you ask questions in this form, you should ask users to choose standard alternative answers immediately after they interact with the functional prototype. Just like the problem with the text description version before. In this way, users can always keep a clear memory and will not confuse this function with other functions in the same survey.
Some people are confused about the order of standard answers in Kano questionnaire. Usually, they don't understand why "I like that" appears before "it must be like that", which seems to be a milder statement.
The logic of presenting the answer in this way is from happiness to avoiding unhappiness. There are some alternative wording for reference, such as:
I like this.
This is a basic necessity, or I hope so.
I am neutral.
I don't like it, but I can stand it.
I don't like it. I can't accept it.
Or this, written by Robert Broth's team:
This is very helpful to me.
This is my basic requirement.
It doesn't affect me.
This is a little troublesome
This is a serious problem for me.
In fact, I think the list of options introduced at the beginning of this guide strikes the best balance between clarity and simplicity. View the link Kano model user guide II (translation)?
It is important to pay attention to how to explain these options, and it is also important to ensure that the respondents understand the purpose of the questionnaire. Choose the best matching answer option set and explain each option to the participants in advance, which will have better effect.
An important supplement to Kano methodology put forward by several teams is to add a question after the combination of functional usability/functional defect. This question asks users how important this function is to them.
Obtaining this information is very useful for distinguishing each other's functions and understanding the functions that are most relevant to users. It provides a tool to distinguish between large-scale and small-scale functional areas, and how they affect users' decision-making on products.
The importance of self-report can be put forward in the following form: "If:? How important is it? " For example, "How important is it if the time to export video is always less than 10 second?" .
The answer should be between 1 and 9, from "completely unimportant" to "extremely important".
If possible, test the questionnaire with team members before sending it to users. If there is any internal confusion about the questionnaire, it must also exist when discussing with outsiders.
Now officially enter the exciting fast lane (I made it up? )。 After tabulating and processing the collected results, we can classify the functions and deeply understand the best method to determine the priority of the functions.
Analysis can be divided into discrete analysis and continuous analysis. These terms are just my thoughts, because these methods lack any standard (or better) terms. They are all references to mathematical concepts and are related to how they draw the results of participants with Kano's classification.
Each method is very useful, depending on the type you need.
The simplest way for us to deal with Kano results is:
Divide respondents according to demographic attributes/role criteria;
Use the evaluation form to classify each respondent's answers;
Count the number of all answers for each function (and demographic attributes) in each category;
The category of the function will be the most frequent answer (i.e. mode);
If the results between categories are close, the following rules will be used (leftmost wins): required >; Expectation > charm > indifference;
If the respondent is asked to provide a self-reported importance ranking (which should be done), please average each function.
Finally, you will get the following table:
If there are multiple results that are not clearly classified, there may be hidden user information that is not considered. In this case, we should go back to the user questionnaire to find clues; Try to find out which users often have the same answers as other users, so as to find "demographic clustering" that may not be discovered.
From the results table, you can sort the functions according to their importance. Since then, the general rule of thumb used in determining priorities is: first, all necessary functions, then add as many expected functions as possible, and finally throw away attractive functions.
This type of analysis is very helpful to get a preliminary understanding, and it is very useful in many cases where more rigorous methods (such as testing design ideas or drawing road maps) are not needed.
Although discrete analysis can help us get started and have a comprehensive understanding of the results, there are still some problems. Namely:
A lot of information was lost in the process. First, from each respondent's 25 answers to one of six Kano categories. Secondly, all respondents' answers are further simplified to a single classification of each function;
Know nothing about the differences in the data;
Give a gentle answer as much as a strong answer. Just think about the difference between "should be so" and "reluctantly endure" in the dimension of attractive function and lack of function.
The following sections introduce a very good continuous analysis method proposed by Bill DuMouchel. Don't worry about these calculations; The spreadsheet attached to this guide has completed these tasks (pull to the bottom and click on the original text to get the data). At present, we only focus on understanding each step.
First, each answer option is converted into the corresponding value within the scope of the satisfaction scale, from -2 to 4. The larger the number, the more the answer can reflect the customer's desire for this function. The importance is also from 1 to 9, as before.
? ? Functions: -2 (dislike),-1 (forbearance), 0 (indifference), 2 (necessity), 4 (like);
Functional deficiency: -2 (like),-1 (must), 0 (indifferent), 2 (endure), 4 (dislike);
Importance: 1 (not important at all), ..., 9 (extremely important)
You might think that the degree of functional defects seems to be reversed. No. The higher the score (positive), the greater the satisfaction potential. For the answer to the lack of function, not liking some functions means strongly opposing the lack of this function. So if it is included, it will have the potential of higher satisfaction, which is why it scores higher.
This asymmetric ratio (starting from -2 instead of -4) is because the classification (reverse type and suspicious type) obtained from the negative answer is weaker than the classification (required type and expected type) obtained from the positive answer. So Dumucher decided to strengthen the positive value.
These scores classify functions in a two-dimensional plane, so there is no need for a standard evaluation table.
The analysis should focus on the positive quadrant, which is the strongest positive emotional answer. In addition, there are other weak answers and the classification of "suspicious" and "reverse" in the table. If a function is "reverse", you can use the skills introduced before to locate it as the corresponding reverse classification, and adjust the scores to the values of "function owned" and "function missing" respectively, so that there is a new classification, biu! Then remove the "reverse" classification.
Side note: satisfaction and dissatisfaction coefficient
When reading various Kano materials, we often see references about satisfaction and dissatisfaction coefficients. The DuMouchel continuous analysis method introduced in this section has a better alternative. But considering how often they are cited, we need to make a brief overview first.
Mike Timko uses "better" and "worse" coefficients to reflect the change of users' satisfaction or dissatisfaction with a certain function. But he didn't name it "satisfaction coefficient" and "dissatisfaction coefficient" in his manuscript, it was just a conventional statement. First, count the total number of answers for a given function in each category, and then apply the following formula to calculate:
Although they do produce numerical results that can be used for relative comparison, there are some problems with these coefficients, which Timko himself mentioned in his article. Most importantly, it faces the same problem as discrete analysis: these values are based on a Kano classification, which is obtained by merging and simplifying the answers of all respondents. The loss of this information will lead to the increase of data instability, and all the answers will be given equal weight, ignoring their original strong/weak emotions.
The scores of function possession and function deficiency calculated by Dumucher's continuous analysis method can also achieve the same goal, and there is no such problem, which is why we focus on it.
If every possible answer is marked with a numerical value, it means that the average value can be used. Here's what we need to do for each function:
1, average score of all functions, function missing, important answer;
2. Standard deviation of function possession, function deficiency and importance score.
According to the average score of function possession and function absence of each function, they can be placed on the two-dimensional classification plane, as shown in the following figure:
Of course, what we are discussing at present is the average score, but it hides the huge differences that may exist in the data. This is why the standard deviation is added to the graph in the form of error lines, so that we can know from the graph whether this classification is in line with or off the target. As shown in the figure below:
The last step is to add importance scores. This dimension can be visualized by converting points in the scatter plot into bubbles, the size of which is proportional to its importance. In this way, it is easy to compare similar functional requirements.
The general priority rules for discrete analysis given before are still applicable here: mandatory >；; Expectation > charm > undifferentiated type. This can be well interpreted as a graphic term:
For small feature sets, another (perhaps better) visualization method is to sort the list through the stack. The table ranks functions in three columns (from high score to low score): potential dissatisfaction, potential satisfaction and importance. In this case, the first two columns are the functional defect score and the functional usability score. As shown in the figure below:
Notice the last two lines. What would you do in that situation? One is an indifference function that has a greater influence on dissatisfaction (but it is actually very close to the necessary type). The other is a function that can obviously improve satisfaction and is considered important by users. In some cases, we need to give priority to one of them. It can be seen that just following some sorting can't solve all practical problems; Many times, we still need to make difficult choices, experiment, measure and iterate.
To see the next section, click Kano Model User Guide (III)-Kano Research Methods (and Toolset).
Click on the complete guide of Carnot model to view the original text.