Learn how to use thematic analysis to strengthen design projects.
As design expands into new domains, we need solid methods to make sense of the world and turn data into actionable insight. I’m on a mission to help fellow designers make the most of qualitative data analysis.
Impact requires solid insight
As designers, we use insight to define user needs, understand human behaviour, and speak for the people impacted by our solutions. But when we talk about insight, we sometimes say things like this:
Only three out of six people talked about it, so it is not very important.
I've done it, and sometimes I still do it. It is human to count, as it is an important way for our brain to make sense of the world. But counting does not always strengthen a qualitative analysis, and counting can make it harder for our conclusions to be heard.
By counting the number informants who talk about a given topic, we may also signal that qualitative insight is insufficient. We open our analysis to a quantitative critique. Listeners who normally rely on surveys with thousands of respondents will discard conclusions based on three out of six people.
A qualitative analysis can du much more. Our aim is often not to count humans, but to understand why they say and act as they do. Our analysis and conclusions may be stronger if we are able to communicate the validity without counting.
In addition, qualitative methods all too often rely on a surface level affinity sorting, without a proper analysis and interpretation. The resulting insight lacks the power to make an impact.
If we aim to use to design to contribute to the big, societal challenges, we need to match our creativity with solid, yet flexible, methods to understand and interpret human needs and behaviour. Thematic analysis is one such method.
The good news: we can learn to use qualitative methods to strengthen our design projects.
An intro to thematic analysis
I use thematic analysis to make sense of qualitative data in design projects. It is easy, fun and flexible. And it works.
Thematic analysis is a systematic approach that can be easily adapted to the context and time frame of design projects. Thematic analysis is widely used in the social sciences, as described by Virginia Braun and Victoria Clarke.
The systematic process has three phases:
- ask good questions and collect data
- code the data using the appropriate strategy
- discover and describe emerging themes
Below I describe each step, highlighting key aspects and pitfalls.
The starting point
As part of a team of designers, I use thematic analysis primarily to understand user needs and context. A key step is to ask appropriate questions to what the data should answer.
Often, my role as a content designer is to help clarify and align the different ideas about what we want to explore at the beginning of a design project. These discussions lead us to a shared understanding, sometimes as a written question or statement as the starting point. We also use this phase to decide how to make sense of the materials we will gather, and if thematic analysis can be an appropriate choice.
Examples of questions where thematic analysis is a good fit:
- What factors are important for successful onboarding of new employees?
- How do people relate to a web page as part of a customer journey?
- What motivates people who use a specific dating app?
- What do expert users need from a digital interface at a high-risk environment?
Using thematic analysis will result in an answer as a list of themes. The themes emerge from the analysis, and gives a big picture answer to each question.
If our questions are more specific to interface interactions, navigation, or conversion rates, we use other methods to collect and make sense of the relevant data.
High quality data
Impactful, actionable insight requires a solid foundation for the analysis. I prefer to use interview notes of in-depth interviews as the raw data.
We rarely have the time to transcribe recordings of interviews. Instead, we do interviews in pairs. One person asks the questions while the other person transcribes the conversation as it happens.
In particular, we try to avoid using short, bullet-point summaries written directly after in-depth interviews. The rapid summaries are based on gut-feeling and personal viewpoints, and often mix anecdotes and interpretation. Brief summaries also introduce bias in the data. One potential risk is that short summaries displace what the informants actually say and mean, and instead add our biased world-views to the mix. If used as data for the analysis, the summaries make it hard to distinguish between what was said and what we think about the topic.
Thematic analysis helps to systematically reduce bias from the analysis, (although bias is never truly absent) and pay close attention to the empirical data at hand. A systematic process can make the journey from interviews to insight more transparent, another factor that can reduce bias.
A flexible coding process
According to Braun and Clarke, thematic analysis is flexible because we can choose between an analysis at the surface level or look for underlying meaning. A second option is to code using a pre-defined set of codes, or approach the data with open hearts and minds and develop the codes as we go.
In the illustration below, the pathways through the coding process are coloured as white and yellow post-it-notes.
In my experience, it is most relevant to use the yellow pathway to analyse insight from the first half of a double-diamond-structured design project. When we want to try to truly understand the reasoning behind actions and choices, we need to go deep into the data to discover underlying meaning. In this way, we can discover unexpected patterns and themes by an open, inductive coding and interpretation.
A code is a short description of the meaning contained in a quote. Braun and Clarke note that it is important to give your full attention to each quote, and then carefully consider the meaning in the data.
Below are some examples of codes assigned to short statements. The quotes are adapted from an interview about how customers relate to a company through their web page.
We try to not be alone in coding the data.We either code in pairs, as a small team, or we code in parallell and then revise and discuss after the initial codes have been assigned.
The coding process depends on interpretation and careful examination, and by discussing the codes as a team we are able to reduce some of the bias in the process.
For an experienced team, coding the transcripts from a forty minute semi-structured interview takes about half an hour. It is worth the time.
There are many ways of organising the process. We mostly use spreadsheets or Miro to organise the coding process and the analysis. The big picture overview and visual representation in Miro has the added benefit of making the visual designers on the team feel at home.
Since most software is searchable, we can re-trace the codes to the specific quotes for the documentation. Also, Miro lets us easily share the process and the results with colleagues and clients.
A fully coded data set may look like this:
The emerging themes
After coding the entire data set, we copy the codes to a new area, and start sorting using affinity mapping.
Affinity mapping is familiar to many designers. A key difference is that instead of sorting quotes or key words from a workshop, we sort a systematically generated set of codes. This helps to keep track of how and when we interpret the data, and makes the process more transparent and reliable.
Below is another screenshot from Miro, showing the progression from an unsorted grid of post-it-notes to fourteen distinct groups of codes.
As we sort the codes into groups, we discuss and explore the patterns and relations that emerge.
A key part of sorting is to start to give names to the groups. We try to use short, full sentences to label the emerging themes, as it helps to clarify what we think the codes in the group mean. Using simpler labels, such as “negative” and “positive” can make sense in some cases, but are rarely sufficient to account for the nuances from an inductive analysis.
The labels are often critical to helps to detect the themes as they appear. By naming each group and discussing them carefully, the themes appear and emerge organically.
The below image shows fourteen groups of codes. It is tempting to stop here, and conclude the analysis with a list of fourteen findings. However, the analysis is not done yet.
The resulting themes
The fourteen groups are adapted from a recent project regarding onboarding of new employees at a large company. We wondered about the following: What factors are important for new employees to succeed?
The fourteen groups did not end up as a list of fourteen findings, but our analysis resulted in three themes to answer the question: employees need a team of colleagues (1), relation between the employee and the manager (2), and digital systems and organizational culture as context (3 A and B).
Each theme consist of sub-themes that bring nuance and richness to the overarching topic. Within theme 1 (affectionately referred to as find your people) the groups include headings like “We eat lunch together”, “We work together in small teams”. The theme describes how new employees succeed through connecting with colleagues both through work and in social situations.
Theme 3A and B are closely related, and describe two different aspects of bigger theme (systems for everyday work and systems for the onboarding process). They could possibly be merged as a larger theme, but from our discussion we decided that the two parts best reflected the data and the interviews.
Communicating the results
We often use storytelling to bind together the different themes as a user story. Listeners can see themselves in the analysis, and relate to the findings in a more human way.
When communicating the above results, our aims were two-fold. First, the three themes provided a clear answer the overall question. The short list was easy to remember and discuss, and it helped make strategic choices on a big-picture level. Secondly, the nuances, quotes, and interpretations provided a clear direction for further work, both as design concepts and new projects.
In my experience, a solid conclusion with two to five big-picture themes brings a human perspective to the discussion and a clear direction for the next steps.
An alternative to counting
It is still tempting to count when presenting qualitative results. But instead of saying that four of five talked about culture at work, we tried to phrase it like this: “In our analysis, we found that the culture at work is important.” or “The informants tell us three important things: A, B, and C …”
Also, a qualitative analysis based on semi-structured, open interviews cannot conclude that a theme that is absent from the data set is not important. Thus, we try to not rank or grade the themes or sub-themes as part of the answer.
In addition, we want to communicate that the analysis is true and accurately represents what the informants shared with us. To do this, we support each theme and sub-theme with representative quotes and discussion. We clarify how we interpret what was said, and how the quote fits to the overall themes.
Developing new design skills
Already during the sorting process, the emerging themes start an enjoyable and contagious creative process. The sentences that describe each group of post-it-notes often provide space for inspiration and reflection, and discussions around the themes can help choose a direction and develop new concepts.
As a content designer, I can also see clear links between qualitative analysis and my writing skills. For example, I use plain language to clarify what we want to explore, to uncover user needs as part of the analysis, to develop relevant and meaningful content, and later to test that the content meets the same user needs. A full circle design process.
A thematic analysis lets us conclude with more confidence, and gives qualitative insight the power it needs for design to increase its impact.
The good news is that you can learn to use qualitative methods to strengthen your design projects!
Virginia Braun & Victoria Clarke (2006) Using thematic analysis in psychology, Qualitative Research in Psychology, 3:2, 77–101, DOI: 10.1191/1478088706qp063oa