How to verify the qualitative character role quantitatively

First, make a simple definition of qualitative research and quantitative research:

Qualitative research refers to the methods of discovering new things from small sample sizes, such as http://www.aliyun.com/zixun/aggregation/8110.html "> user interviews, usability tests, focus groups, etc."

Quantitative research is a method of testing and proving certain things with a large number of samples, such as questionnaires, website traffic/log file analysis, etc.
Then why do we have to do quantitative research after getting qualitative personas? The aim has the following two:

1, we use a large number of samples to verify the classification variables and the types of personas in qualitative studies.

2, access to qualitative research (small-scale sample size) when the data can not be obtained, such as the user's product use, demographic indicators, can help the product side more convenient to identify various types of users, judge the level of activity of various users, Jambi, etc.

So, when should we do quantitative validation? Depending on the circumstances:

1, when the qualitative study is completed, when it can be carried out;

2, the quantitative study can be done immediately after the type of qualitative personas is obtained, as shown in the following figure:

Finally, how do we do quantitative research to validate qualitative personas? The following will be shared with the questionnaire after the analysis of the general process of verification and some of the experience:

The 1th step, questionnaire survey: Access to data

The questionnaire needs to measure the following two parts:

(1) The classification variables (a,b,c ...) obtained in the qualitative study are used to validate the classification variables and the character role types. The so-called classification variables are the key differences that we use to define personas in qualitative studies (their specific goals, behaviors, and perspectives). We need to design observational variables (a1-an,b1-bn, C1-CN ...). To each category variable (A,b,c ...) For measurement, each observation variable corresponds to a questionnaire question or option. The observational variables are mainly designed according to qualitative research, so the researchers of qualitative research should be involved in this process.

The following is an example of a consumer's satisfaction with a company's business to illustrate classified variables and observation variables:

(2) The use of products, demographic indicators and other variables, used to obtain qualitative research (small-scale sample size) when the data is not available, can also be used to classify the categories after the description.

For the use of products, we will generally consider the following variables: Service life, Login frequency, functional preferences, etc.

For demographic indicators, we will generally consider the following variables: Sex, age, education, occupation, industry, income, etc.

In this step, the product staff may mention a lot of their concerns about the product issues, we need to choose according to the situation, there are some issues that are not suitable for the quantitative validation of personas, such as whether users will use shortcuts, and so on, to communicate with the product personnel for future research to be resolved

The 2nd step, to the variable clustering: the validity of the observation variable test

This step is the basis of the subsequent analysis to answer this question: These observational variables (a1-an,b1-bn, C1-CN ... Can be effective on the classification of variables (A,b,c ... ) for measurement?

The solution is: to observe the variable (A1-AN,B1-BN,C1-CN ... As a cluster variable, the variables are clustered, if the clustering results of the variable classification (1,2,3 ...) and classification variables (a,b,c ...) ) is basically consistent, that is, A1-an clustering for variable classification 1,B1-BN clustering for variable classification 2,C1-CN clustering for variable classification 3, and so on, it shows that these observation variables can be used to measure classification variables.

It is important to note that although there is a more rigorous and scientific approach to the reliability and validity of the measurement, the complexity of the operation and the actual value of the product are only used to test the validity of these observed variables.

3rd, user clustering: Validating personas

This step validates the persona by answering the following two questions:

(1) Using these classification variables can be divided into several categories of users? Are the types of categories and personas consistent? --The solution is: using classification variables to cluster users

(2) which variables in these classification variables can effectively distinguish between several categories of users? The-– solution is: to examine the significance of differences in different classification variables

When using classification variables to cluster users, there are two options:

(1) The observed variable (A1-AN,B1-BN,C1-CN ... As a cluster variable, the user is clustered

(2) Classify variables (a,b,c ...) ) (need to be observed variable (A1-AN,B1-BN,C1-CN ...) ) as a cluster variable, the user is clustered

Both methods can be validated by the final user classification to verify qualitative personas, and the difference is that the 1th method is more suitable for different classifications in each observation variable (A1-AN,B1-BN,C1-CN ...). ; The 2nd method is more applicable to want to understand different classifications in the classification variable (a,b,c ...). ) on the situation of the difference

To explain, the difference between "clustering" and "clustering" in the previous step is that the objects of clustering are different, "to the user clustering" is through the classification variable to divide the user into several kinds, the cluster object is the user; "To the variable clustering is to classify the classification variable to several kinds, the cluster object is the variable.

In the analysis of the resulting user classification, some of the findings may be found to be different from qualitative research, which can be divided into the following 3 cases:

(1) To get the type of persona not in the previous qualitative study. This is a very likely scenario, because it is easy to recruit users who use a deeper product and are very loyal to the product, so that we may overlook certain types of users, such as the use of products with a lighter, less motivated user. Such users for the product is generally less important users, in the report can be described;

(2) There was no type of persona in the previous qualitative study. This may be a biased sample of the questionnaire or in the user community, the number of users in this category, should be based on the importance of such users to the product to make trade-offs;

(3) A character role type in the previous qualitative study needs to be split into several categories. If in the quantitative verification found that several categories of users do have a large difference in the classification variables, for the product is belong to several types of users with different needs, then need to split into several categories of users.

In short, whether or not you need to delete a certain type of users, are based on the importance of the product to determine the user. If the product is important, it can be judged by comprehensive qualitative research, quantitative research and discussion with product personnel.

The 4th step, the description of various types of users

After 2 or 3 steps, the validation of qualitative personas has been completed, and several categories of users can be described using variables such as demography and product usage. To ensure that the sample is unbiased, this step can answer some of our questions as follows:

(1) What type of user is the highest in the user? --The solution is: calculate the proportion of the number of users

(2) How to identify several types of users from demographic indicators? --The solution is: to describe the population attributes of all kinds of users, to calculate whether there is a significant difference in population statistics

(3) How to determine what kind of users are our active users? --The solution is: to calculate the use of products or important functions of various types of users

(4) How to determine what kind of users are our important users? --The solution is: analysis of the functional use of various types of users or functional requirements, combined with qualitative research and communication with product personnel to judge

。。。。。。

These problems may be some of the requirements that product personnel emphasize when doing qualitative research, and quantitative research can help us find answers with a lot of data.

PS: a misunderstanding:

When communicating with a product person on a persona report, product personnel often have the following misconceptions about personas:

Is it possible for everyone to correspond to a particular persona? I do not know what kind of myself, is not everyone is a mixed user?

In fact, although qualitative personas are validated by the classification of users, personas do not fully correspond to real users. Because personas are only a prototype, a certain type of persona has some common behavioral patterns among real users, while a real user may have a pattern of behavior represented by certain types of personas.

Source: http://uedc.163.com/4876.html