How to Create Personas Using Grounded Theory

Over the past eight years, I have slowly adapted my academic research habits to work in various product and tech companies. However, the fast-paced nature of companies and, thus, user research has always been a struggle.

Just like in academic research, we cannot simply get a topic and dive right into the research. There has to be time to think about questions, question questions, and create the most relevant and holistic approach.

This is also the case when it comes to creating large-scale deliverables. The times I've struggled the most are when colleagues or clients expect personas in a matter of a few weeks.

For a while, I applied some short-cuts to get personas or reports out faster. I imposed pre-set theories or codes on the data to make it easier to synthesize. For example, when I was (quickly) creating personas for a company, I decided to use the following categories on the personas before I even looked at the data:

1. Goals
2. Needs
3. Pain points
4. Motivations

When I went through my data, I looked for those particular ideas, highlighting each. While there was nothing wrong with this approach, I started analysis with a pre-set idea of the data I was looking for rather than finding and generating tags.

This method was more straightforward and efficient than finding tags and categories after I sifted through the data, otherwise known as inductive research.

But when I finally had a persona project that I was able to dedicate ample time and resources to, I decided to go back to my years of academic research and use a grounded theory approach to build the personas.

What is grounded theory?

As a UXer, you’re likely to have done some form of grounded theory before even if you’re not familiar with the specific name.

In the 1960s, two sociologists, Glaser and Strauss, founded the theory. Like its name suggests, it’s about grounding your theory in empirical data.

Instead of starting with assumptions or hypotheses, you start from scratch and create your theories after starting your research. For example, if I go out and buy 20 plants and all 20 of those plants die, I generate the theory that I suck at gardening.

This theory emerges from the data, rather than me theorizing that I suck at gardening and then going out to find ways to prove that theory.

Instead of starting with a hypothesis or assumption, you go through the following steps:

1. Define a concept or topic to learn more about
2. Collect data from a variety of methods (ex: 1x1 interviews, contextual inquiry, diary studies, secondary research)
3. Code/tag the key points that emerge from the data
4. Group the codes/tags into similar concepts (clusters)
5. Form larger categories
6. Define the core category
7. Build theories from the categories

What does it have to do with user research?

Grounded theory is a fantastic way to explore the problem space or conduct discovery research. With GT, you aren't trying to answer a particular problem but are exploring a broad topic where little existing knowledge lies. GT is also a fantastic way of understanding processes, answering the "how" questions of qualitative research.

As with any framework, there is a time and place for grounded theory:

• When you need to understand a broad topic or process and want to immerse yourself in the data
• When there is little existing knowledge on the subject or process, either by previous internal research or external research sources
• If you have time to sit with the data and don't have pre-existing theories or hypotheses
• If you collect a large amount of qualitative data (you can also use this to inform a follow-up quantitative study)

Just keep in mind that the process of GT can take a considerable amount of time. There is ongoing recruitment, research, and analysis, not including time when you need to sit and think. If you have this luxury, go for it! Grounded theory is a rewarding method.

If we take user research and interlace it with grounded theory, we can understand broad topics or processes we know little about. For which deliverables would this type of research be helpful? Personas, journey maps, mental model diagrams—all of those rich, qualitative visuals that provide colleagues a deep understanding of humans.

Using grounded theory to build personas

I have used grounded theory a few times to build personas, and they were better than any template or approach out there. Here is how I conducted this research:

Context

I worked for a company that helped other companies gather user-generated content onto their websites. User-generated content (UGC) is when you see Instagram photos of people just like you on websites, showing off beautiful images with their purchases from that company.

A typical example is interior design stores. You might see pictures of people's beautiful homes with products from the store. This is a way of advertising for the company and gives visibility on social media to the customer (more likes, follows, etc.).

Define the topic

Companies would buy our platform and then, we would gather the UGC for companies to comb through and pick the best photos to use.

There were several different people involved in the end-to-end lifetime of this product, and we were unfamiliar with most of them. So we decided to first focus on the "front line," those people who were in the weeds with our product, using it every day.

Our topic was better understanding our primary day-to-day users, social media managers. Our expected outcome was a persona (or two) that represented this role.

Collecting data

We started with stakeholder workshops, where we pulled together colleagues to explore the biases and assumptions we held about social media managers. So, we spoke to account managers, sales, customer support, product managers, and marketing to begin to weave together a proto-persona.

This proto-persona was filled with anecdotes and experiences from these departments and helped us understand what we thought we currently knew about this role. This approach was a great starting point because it allowed us to take a step back from our biases and ensure we conducted sound research.

Since we wanted to understand our users more deeply, we decided on a mix of 1x1 interviews and contextual inquiry. We also triangulated data from different sources, such as customer support tickets, account managers' experiences, and relevant product analytics.

Since this was a generative research effort, we initially interviewed and observed 25 social media managers to ensure we reached saturation.

Coding and tagging

Now, this is where the linear path starts to break down. Coding and tagging are words that represent something you found in your data.

We use codes/tags to categorize raw research data. This approach allows us to identify themes or patterns between participants.

A lot of times, user research occurs in this trajectory:

1. Create a research plan
2. Do all the research
3. Come back and analyze it
4. Share it

Using grounded theory means using a "zig-zag approach" to data collection and analysis. This approach means that you do a bit of research, come back and analyze, do a bit more research, come back and analyze, and on. The "zig-zag" approach allows you to continuously refine your codes/tags and, later, clusters and categories.

I employed the zig-zag approach to these personas. I went out and did about two to three interviews and would come back to comb through the data to create tags. I used the inductive method of analysis, where I didn't make any codes until I went through some of the data.

When I was looking through the data, I used open coding. I looked at the data for interesting or meaningful information. The way I define "interesting" is when the data:

1. Would be important to teams who need to make decisions about the future of the product or features
2. Made me go, "huh, I had no idea"
3. Surprised me because it went against current hypotheses

After my first few interviews, I went through the transcripts and started the open coding process. I looked for meaningful data and then generated a theme behind it.

For example, in the first few interviews, the participants had expressed not having the capacity (time, resources) to comb through all the UGC our platform collected continuously. They also talked about a high level of irrelevant or unusable content in the platform.

I went back out into the field, continued to collect data, and constantly analyzed it.

Group the codes/tags into similar concepts (clusters)

As I mentioned, I noticed some themes emerging in the data, such as the lack of capacity and the irrelevant content. Participants also noted having a challenging time understanding which metrics they should be tracking.

I then started to create these concepts and bring together similar data. For example:

Data

P1: "I never have the time to go through all the content. I'm constantly getting pulled in different directions and can't sit down to comb through all the data."

P2: "It's impossible to get through all the content because I don't have time, and everyone wants the UGC so quickly. I just don't have the bandwidth!"

With this, I tagged these data pieces with "lacking time/resources," which became my first concept. I then saw more quotes come through about this concept, so I continued to add them.

I also did this with the metrics data I collected, which appeared through many interviews. This became a new concept called "uncertainty behind metrics."

Another example was with the data surrounding irrelevant content. Participants mentioned this problem many times, so I created the concept "irrelevant content."

Form larger categories

Of course, I had many more concepts from the data, but we'd be here forever if I went through them all!

Once I had identified concepts, I started to form categories. These categories are filled with similar concepts. For example, if we look at the two concepts of "irrelevant content" and "lack of time/resources," these feel very similar to each other. Whether it is a lack of time or too much content, the sentiment is about feeling overwhelmed on the platform. This could also include the "uncertainty behind metrics" concept we saw earlier.

We could then roll all of these concepts into a category called "Feeling overwhelmed by the platform."

I used the example of only interview data, but other data should go into the categories, such as from observations, customer support tickets, surveys, or analytics.

Remember to compare constantly

As a nod to the "zig-zag approach," I constantly compared the concepts and categories with any incoming new data. This process helped ensure that I was still going in the right direction and helped me further hone my concepts and categories.

Constant comparison of new data also helps indicate when you've hit theoretical saturation. Theoretical saturation means that we are no longer learning anything new during the sessions, and no more relationships are being made during the analysis.

Define the core category

Once I had immersed myself in the data and met theoretical saturation, it was time to move to the next step. I pulled all my concepts and categories together and used selective coding to create the core idea from the research. The core category is the basis for defining the theory.

So, in my research, here are some of the categories I found:

1. Overwhelmed by the platform
2. Uncertain of how to make progress
3. Feeling stressed by expectations from colleagues
4. Prove ROI of social media

I wanted to create a core category to link these together so the team could understand this persona deeply. After much thought, my core category became:

Social media managers are overwhelmed with the amount of work they need to do and rely on our platform to guide them to do their job more effectively and efficiently.

Define the theory

This step is where academic research takes a right-hand turn, and user research goes left. Instead of building a theory, we need to create deliverables such as:

• Insights
• Jobs to be Done
• Personas
• Journey maps
• Mental model diagrams
• Problem statements

The core category above became the tagline for the persona. I then used the other categories as data in the persona, including examples from the original concepts. Since there were quite a few categories, I had to choose the ones that came up most often and felt most impactful to the participants.

The persona included the major categories and supporting evidence from relevant concepts. It gave teams a great understanding of this role, beyond the typical pain points, goals, and motivations.

Using grounded theory in user research is a fantastic practice that can open your eyes to a new way of analysis. Although it can take time, the reward is valid, reliable, and in-depth knowledge that a team can use for years ahead.