The State of User Research 2022 report has arrived! Explore the data.
SUBSCRIBE TO OUR NEWSLETTER
March 31, 2022
Drowning in textual data and unsure where to begin? This analysis method can help you make sense of complex qualitative datasets.
Qualitative coding is the process of categorizing, labeling, and organizing qualitative data in order to identify themes and analyze relationships in the data.
Coding is also helpful if you want to slice your data by different segments (such as ICP, roles, sentiment, etc.) to identify patterns and themes.
Here’s a qualitative coding example from the open-ended response analysis of our State of User Research 2022 Report:
As you can see, we distilled each unique response into a uniform “code” or theme to better understand the qualitative data we collected. Although nearly a third of participants said they don’t have a method for tracking the impact of user research, a significant number of participants said they used methods like analytics, stakeholder feedback, engagement with research artifacts, and more.
For example, you could create codes like “1 - Career” for open survey responses related to participants’ careers or “2 - Family” for diary study entries about the participants’ families.
In the initial stages, these codes can remain fairly broad, but as you dig deeper, you can also assign sub-categories (e.g. “Career - hiring” or “Family - extended”).
As a general rule, though, you shouldn’t get too granular with your codes to avoid confusion (for example, “Family - extended” is ok, but adding additional codes for first cousins, second cousins, third cousins, etc., will likely become too cumbersome for practical use).
When you notice qualitative data that aligns with the themes you identified, tag it with the code.
If applicable, you can assign more than one code to the same data (for example, a survey response that says “I love working remotely because it gives me flexibility over my schedule and working environment, helps me work more efficiently, and leaves me with more time to spend with my family” could be tagged with multiple codes like “Remote Work,” “Flexibility,” “Efficiency” and “Family.”)
For example, if 25% of the responses were tagged “Career” and 50% were tagged “Family,” you can conclude that family-related themes are more prevalent in this particular study.
You’ve now entered the actual analysis stage of qualitative coding.
Create categories for similar types of codes (e.g. “Laptop,” “Office,” “Salary,” and “Resume” could all fall under the category of “Careers”) to organize your data further and discover new connections between data points.
As you cross-analyze codes and categories against each other, you can begin to identify themes and develop a narrative about what your data means.
This is the general approach for qualitative coding, but it will look a little different depending on which type of coding you choose to use.
There are two types of coding in qualitative data analysis: Inductive (a.k.a ‘ground-up,’ ‘concept-driven,’ or ‘open’) and deductive (a.k.a. ‘top-down’ or ‘data-driven’).
Inductive coding is an approach to qualitative data analysis which creates codes from scratch, based on the data you’ve collected. Inductive coding is an iterative, time-consuming process, but it provides you with a thorough and unbiased understanding of your data.
Inductive coding is typically used for the first round of analysis, or in exploratory research, when you’re analyzing a particular type of data for the first time with no prior expectations or set measurements. In this approach, you read through your qualitative data and assign relevant codes as you go along. You’ll probably need to review the data multiple times to make changes where codes need to be split, combined, removed, or created anew.
Deductive coding is an approach to qualitative data analysis which uses a predetermined set of codes, making it a quicker and easier way to analyze qualitative data. (Usually, though, these predetermined codes come from an earlier inductive process).
Deductive coding typically works best for later stages of analysis, such as in evaluative research.
For example, you might start out with a set of themes you’re interested in analyzing: Analytics, Stakeholders, Engagement. As you look through your data, you assign these codes wherever they appear.
Although deductive coding is quicker and easier than inductive coding, be mindful of bias. With a predefined set of codes, it’s easy to become blind to other major themes that may come up in your actual data set; don’t focus so hard on proving your own hypothesis that you miss them.
Often, researchers prefer to use a hybrid approach of both deductive and inductive analysis; this way, you ensure that your pre-defined research questions are answered while remaining open to any new themes that might emerge along the way.
You can do quantitative coding using Google Sheets, Excel, or—if you’re a glutton for punishment—by hand. But if you’re handling an especially large dataset or simply want to streamline the process as much as possible, here are some great automatic coding tools you can try:
If none of these look like the right fit for you, check out The 2021 UX Research Tools Map for more. (P.S. Our 2022 edition of the UX Research Tools Map will be released later this year—keep an eye out for the updated version!)
The data that comes from qualitative research—quantitative research’s ‘touchy-feely’ counterpart—is sometimes dismissed for having a higher risk of bias and subjectivity.
Using qualitative coding in your approach to analysis, you can standardize the quantitative data you collect, limiting (but not eliminating) bias and improving both the reliability and the replicability of your insights.
For more information on analyzing UX research, check out the Analysis and Synthesis chapter of the UX Research Field Guide.
Content Marketing Manager
Marketer, writer, poet. Lizzy likes hiking, people-watching, thrift shopping, learning and sharing ideas. Her happiest memory is sitting on the shore of Lake Champlain in the summer of 2020, eating a clementine.