User Personas: How to Create Them - Part 1 - Review

This article on User Personas is based on the learnings from CXL Institute.

Let us start by defining User Personas.

What are User Personas

CXL Institute puts forth the definition by Alan Cooper as:

" A precise descriptive model of the user, what he wishes to accomplish and why".

The main words in this definition are - Model and Why.

Model can be said to be an approximate description of the User.

Why is the main core that describes the user's motivations to buy.

Two misconceptions regarding the user personas is that they are not of good quality and they take a lot of time.

Also, many feel that for a high quality user persona, one needs to invest considerable amount of time.

It depends on what you put in to get a user persona.

If we put in good data, we can get good user personas.

One thing that makes user personas bad is irrelevant data.

So, one has to ensure that relevant data is put in to create the user personas.

Another key element of a bad ingredient for user personas is data that is not actionable.

To create good personas, the data needs to be actionable.

For example, let us see the example of a Car Shopping Survey.

Example of Irrelevant data or question:

What is your experience driving a Car.

The reason for this question being irrelevant is that it talks about the data in the past.

Also, it is not going to necessarily connect the consumers to the car buying application that you plan to develop.

Example of Relevant data or question:

What is your experience buying a Car.

Similarly, we can see examples for Actionable or not actionable.

Example of Unactionable: What is the MPG on your current car.

Example of Actionable: What is your desired MPG on your future car.

Steps in making User Personas: Step 1

Making User Personas can be broadly divided into three steps:

Step 1: Collect the data

Step 2: Identify the groups.

Step 3: Build the Archetypes.

Personas are only as good as the data.

The data is the foundation.

For this, the questions need to be of good quality.

Three qualities of good questions are that they are relevant, actionable and unbiased.

We have seen the examples of relevant and actionable questions in the previous section.

Here, we discuss one example related if the question is biased or unbiased.

Example of a Biased question:

Would you rather buy a car from a dealer or a stranger?

The words like dealer and stranger give a biased outlook to the question.

The example of unbiased question would be:

Would you rather buy a car from a salesman or from a private seller?

These questions are usually part of a Survey.

One approach is to lump the questions together to make the flow easy.

The features can be put on the left most column cells and the options of least likely to most likely can be on the top of the following columns.

It is advisable to put words like least likely or least important to most likely or most important.

The reason being that only numbers from 1 to 5 will most likely confuse the survey takers what 1 means and what 5 means.

If any participant assumes 5 being the worst or least important, the data analysis process can get affected negatively.

In Step 1, it is also important to know and understand the process of building the survey.

We can use a good number of tools available in the market.

Few of them are paid, while others are free of cost.

Tools like Qualtrics is paid and high price, but it has high flexibility on building the survey as it has a lot of options and features to explore.

Another tool is Survey Monkey. It has a free version and a cheaper paid version.

Another survey tool I know of is Survey Sparrow. It has very good features.

I have covered a separate article on Surveys and features available on Survey Sparrow.

One of the free options available for surveys are the Google Forms.

Downside of this is that we have limited functionality in terms of the type of questions hat we can frame for the survey.

While selecting a tool, it is important that we also see that the tools allows asking open-ended questions.

The open-ended questions capture a lot of information about how people feel about a certain thing.

Another key step in the collecting the data is recruiting respondents.

One way to do it is surveying the users on your website.

The tool which you can use for it is hot jar.

The way hot jar works is in the form of a snippet.

You have the control to activate the surveys - turn it on or off.

Another way is to survey the general public.

You can do this in marketplaces.

One of them is Amazon mechanical turk.

On this platform, people are paid to do specific tasks online.

Amazon mechanical turk is widely used in academic world.

It is also used in the industry.

The users of this platform also called the workers provide the required hits for the survey or cluster analysis.

The Participants - How many are enough

We have to try to get as many participants as possible.

Preferably in the range of 300 to 1000 Participants.

But if you are not able to get these many participants, you can try to manage with 100 plus participants.

One key point here is to make a pilot test before going big.

You can select a few participants from 10-50 depending on your budget.

Then use this data to check if the questions need reframing or better words can be used.

One more thing which can be done is to work collaboratively after the pilot test survey is completed.

More like a peer review.

You can share the preliminary results with your colleagues and ask them to peer review the whole survey or any specific part of the question.

The topic hopefully will be continued in another article.