User Personas: How to Create them Part 2 - Review

This article is a continuation of the topic of User Personas.

The article is an extract of the understandings and learnings from CXL Institute.

In another article on this blog we started discussing the topic.

We discussed what are user personas, the steps to make user personas.

In particular, we discussed Step 1 in detail, which is to collect data.

In this article, we will continue the discussion of Step 1 and try to discuss the Steps 2 and 3 reach a conclusion.

Step 2 is identifying the groups in the data.

Step 3 is building the archetypes related to user personas.

In collecting the data, it is also important to simplify it.

The data that we collected has some purpose.

We should be able to extract value from it.

One has to know how to extract value from a survey.

The questions and answers of the survey should be translated to a select key factors or issues that you want to touch with the survey.

This leads to how to cluster those responses around those key factors which eventually leads to the personas.

One common form of file that is available with most software is CSV or excel form.

One can download the data from the survey in this form.

Organize the data

Once we have the data in a CSV form, we can now organize the data to start making sense out of it.

We can label the rows as Respondents.

Column refers to the answer to a particular or specific question.

An organized form of data is also called the Tidy data.

When we start to organize, we tend to realize that we have collected a lot of data.

Let's say we had 16 to 20 data points or questions for each person.

The key question here is how do we form clusters around this data.

How those clusters can then be translated to personas.

We simplify the data by using a method called Exploratory Factor Analysis or EFA.

The theory or main concept behind this method is that the 16 to 20 questions have answers to few key underlying factors.

For 16 questions in the survey, there may be only 3 issues or factors.

It becomes easier to see clusters forming from these key factors.

It is not that we need to know the factors before designing the survey or what questions to be framed.

Sometimes the respondents respond to the survey questions in a way that is totally not anticipated by us.

This is one of the advantage of exploratory factor analysis.

The advantage being that as a survey designer we are able to group the questions based on the survey responses.

We can see how the responses are made to the survey questions.

Identify the Factors

When we start identifying the factors, a good deal of software ask us to put the number of the factors beforehand.

We cannot know the exact number of factors beforehand.

Sometime it can be 3 or4 or 5.

That is the reason it is an iterative process to identify the factors.

The next thing would be to decide how much of influence each of the factors that are identified have on each of the questions.

In other words, how much information to the factors is being conveyed by the answers to that questions.

The influence of the factors on the questions can be highly varying.

Each factor can vary one question each or each factor can influence multiple questions.

Sometimes even all the questions.

This is the main job of the factor analysis.

To decide how much influence each of the factor has on each of the question in the survey.

Interpreting the Factors

Let us take the example of Car buying again as taken in Part 1 of this article.

For the car buying experience, the factor Value is of importance which can have words like Value, condition (of the car), and accident history of the car.

Here, the theoretical factor is Value.

All the three above are related to the factor Value.

Another theoretical factor can be expert opinion.

This can include further sub factors like reviews ratings and expert recommends.

Based on the respondent, the factors can be given a high or a low value.

Now, not all factors are interpretable in the first attempt.

There can be factors a little different.

The third theoretical factor that we can discuss in this example is fuel/ environment/ safety.

Along with the influence of the factor on each of the questions, we also assign a weight of each of the factor for the respondent.

In this way, we can get the factor scores.

Based on the factor scores, we can then understand as to why a particular respondent gave that much weightage to that question.

This is another key advantage of factor analysis.

It reveals the underlying why based on the two theoretical constructs of factors and the questions.

The Working of Cluster Analysis

With the factors scores for each respondent, it becomes clearer on how the clusters are getting formed.

Based on the scores and factors, we can now group the respondents in to cluster or groups.

To represent a cluster analysis, we can use different visualization tools.

One way to do it is in three-dimensional axes.

If we have three factors, we can assign each to an axis - x-axis, y-axis, and z-axis.

The scores can form spherical shapes.

There can be respondents that can share the responses on each of the factors.

Let's say we have given colors red, blue and yellow to the factors.

So the points can show what importance each respondent has given to the red factor, the blue factor and the yellow factor.

Identifying the Clusters

It is advisable to use a computer program or software so that we can easily scale the responses.

One thing we can do after getting the scores and identifying the clusters is take an average.

We can average the cluster members.

We can do this by averaging the cluster scores for cluster member to find the average cluster member.

The topic is highly interesting and hence hopefully will be continued in another article.