Beyond Post-Its and Affinity Mapping: Qualitative Data Analysis
We all know this: A wall full of sticky notes with observations and quotes from research that get summed up in clusters. For a usability test, this is straightforward and if you’re looking for some inspiration for quick idea generation, this might be all you need. But sometimes, you might wonder if you missed something crucial within all theses notes. And for upfront, generative research in the problem space with many interviews, the wall quickly becomes overwhelming — or just too small.
Where do you sit in the research continuum?
There are different levels of research depending on your research questions and goals. If you are in new terrain, you will need to explore and be open to learn something new. If you are already familiar with your topic, you will have a lot of hypotheses and rather look for confirmation. Speaking of different research levels, evaluative user research will be more on the right sight of the research continuum while generative research tends to the left side.
Generative user research at the front end of innovation is gaining new understanding to people’s behaviors, meanings and needs. As we humans are irrational, and usually people cannot state their number one problem for you to solve, we need to read between the lines and draw our own conclusions. Analyzing and reducing the data to display it in an organized way helps to cut through the clutter. At the outcome, frameworks like 2x2 matrices, venn diagrams, maps etc. help to give an overview and visualize the relationships between different categories. Also, on this level, the synthesis that came out of the qualitative data is better prepared to connect with desk research or quantitative research.
Reflecting on your assumptions
Disciplines like social science provide some methodologies on how to dive deeply into your qualitative data without getting lost. The tricky part is to balance your open approach with already existing hypotheses as well as your own bias and cherry picking. No qualitative research can happen in a black box, so at first you should be clear about your pre-existing assumptions about your research topic, users, etc. This should have already be done at the very beginning of the research and is reflected in different sub topics and questions in your interviewer guideline.
No qualitative research can happen in a black box.
There are many different approaches to analyze your qualitative data given the different research disciplines. In the guide “Lives & Legacies” Ping-Chun Hasiung breaks down a general approach which takes into account working with your pre-existing assumptions and searching for new ones. On the Lives & Legacies website, you can go through interview scripts and examples of data analysis and try it on your own.
The four lenses for structured analysis
So, when you are going through your notes and interview transcripts, there are different lenses for analysis: open coding and focused coding, descriptive coding and analytical coding. You are annotating and summing up your transcripts, resp. notes, by putting short labels/codes to it. Open coding looks for new topics while focused coding scans the notes for specific topics that you already have had identified in advance or through open coding. That is, focused coding is done at best in a second, separate round. With descriptive coding, you take a “neutral” stance — think of a crime scene observer from the police. Analytical coding opens the door for interpretation.
There are different lenses for analysis: open coding and focused coding, descriptive coding and analytical coding.
While descriptive coding finds the different recurrent patterns, analytical coding asks what the different patterns mean and under what conditions they have occured. You also look for the grit in the gears, like extreme positions, surprising observations, contradictions or tensions between interpretations. Here, you go beyond describing the grit as well as different types of variations and interconnections: What do they mean?
Probably, like with lots of “human stuff”, complexity comes into play and it is “both x and y”. There is no right and wrong and different researchers can come to different conclusion. This is why a research team is invaluable for qualitative data analysis. If not done in advance, this is also the time to bring in your knowledge from the research and design community. For example, when I was working with research data in a project for the BBC on news consumption, I took categories from the established uses and gratification approach in communications, which provides insight into different reasons people use media. What about your research? Maybe the jobs to be done approach hints to meaningful categories?
As a starter, looking for the following general phenomena might help to find patterns and codes in qualitative data:
- Process
- Activities
- Events
- Relationships
- Strategies
- Perspectives
After spending some time with your data, you develop plenty of categories and sub-categories and then might merge some of them. It is hard to tell when to stop analyzing. If you feel you cannot detect and learn something new, it certainly is time.
It is hard to tell when to stop analyzing. If you feel you cannot detect and learn something new, it certainly is time.
Digital tools to cut through the clutter
Analyzing lots of text and observations can become messy. Researchers use spreadsheets or special software for coding and retrieving data. It helps you to draw the connections, order or recode your categories, and get an overview of the quantity of codes.
Coming from social science, programs like MaxQDA, nVivo, atlas.ti, dedoose or f4 analyse are used by researchers. Some of these also integrate other media than text as input. They also help to organize your thinking process with research notes. In the UX and design research context, tweaks of the intelligent spreadsheets built on Airtable are gaining momentum, with Aurelius as a specialized new tool inspired by the latter.
