Open-ended survey questions often provide the most useful insights, but if you are dealing with hundreds or thousands of answers, summarising them will give you the biggest headache. The answer lies in coding open-ended questions. This means assigning one or more categories (also called codes) to each response. But how does one do that?
Two key approaches to coding open-ended responses
Manual coding has been invented many years ago in qualitative research. Here, a person codes each free-text response with one or more category. If you are dealing with hundreds or less responses, you can do it yourself, or you can hire an agency.
Automated coding has become popular in recent years. This is an area of text analytics, where algorithms are used to do this task. Data scientists could use an NLP library, but tuning such libraries is hard, and the results are often difficult to interpret. Solutions like Thematic simplify this process.
Best practices for open-ended coding
Whether you go for the manual or automated approach, it’s a good idea to learn best practices from people who have been dealing with text for decades. Make sure to read our guide to open-end coding will help you learn how manual coding works.
The main recommendations are:
Fight Bias: Don’t start with a conceived notion of what you are hoping to find.
Provide Coverage: Make sure that the code frame covers a wide range of contrasting categories
Ensure Consistency: Keep iterating until each customer response is tagged with all themes present in the dataset.
How to use coding results to make sense of customer comments?
Either you will use an existing dashboard, or you create one yourself using Excel, Numbers or Tableau.
Here are the key things you need to capture:
- What is the overall importance of each category in your code frame.
- How do these change depending on the segments of data
The best chart to capture this, is the old good bar chart. Here is an example of what it could look like in a custom dashboard:
And here is an example of how it could look in Excel or Numbers:
Of course, what it looks like is not as important as whether the analysis is accurate and useful.
