Secrets of qualitative data analysis revealed
Secrets of qualitative data analysis revealed
Interested to know the secrets of quick and easy qualitative data analysis?
Got 300 comments from your survey and dreading having to read them all, three or four times?
Well here they are!
[for now*] Human beings remain the fastest analysis tools for qualitative data.
But don’t despair. The following is a description of the process you can use to shorten the process as much as possible. And keep reading, in other posts I’ll discuss how estimate the time you will need to analyse your data.
SO, what’s the process. Here it is:
The details:
1. This first step depends on how many responses you have to analyse. Reading all of them, even if you are going quickly, might not really be feasible. So, what do you do? Select a sample of your sample. The absolute minimum you should look at is 30 from a variety of respondents. Try to look at 10-15% of your responses if you can.
The point of this review is to find the key themes that seem to be most frequently repeated. This is where you decide what your themes will be, before you code the rest.
You may want to get Excel out at this point, or get yourself some kind of spreadsheet. Our team always makes a table, with each respondent, any details collected about them, their comments, and then columns for codes and sentiment.
You can also some software tools to help you find common words in this stage, like a word-cloud generator, but you will still have to read the comments. At best software can just point you in the right direction.
2. Once you have your 5 to 6 key themes, you should check them with someone you are working with. This might be your client or someone internal to your process. You’re looking to see that your initial review makes sense and that it uses language that is appropriate for the context.
In ‘most’ cases, 5 to 6 themes will be enough, any more and the analysis will be difficult for your audience to take in. You can also use an ‘other’ or ‘misc’ category, if you have 3 or 4 very common themes, and there many many others that aren’t significant enough to make a category on their own.
3. Check your draft themes with your client or colleagues. Sometimes there is particular language that is particular to the population involved in the study, it doesn’t hurt to make your analysis consistent with the terms that they commonly use and have a join understanding of.
4. Return to your qualitative data and add your codes. You may also be adding sentiment at this stage (choose your scale of sentiment appropriately for the context). You may have more than one code for each comment, in this case consider how you will report that before you go down that path.
5. Finally you should end with a a full table with comments, codes and sentiment, with some details about the respondents (if you collected them) and you can make a report. Reporting themes and sentiment visually on the same chart can be an excellent way to show not only how regularly something is mentioned, but also whether there is broad agreement or disagreement between respondents.
And that’s it. Of course there are great advancements coming along with software and machine learning that one day will be able to do this for us, but for now humans are the fastest means of comprehending accurately the nuance between many respondents’ answers.
Keep reading as I will do a test of various artificial intelligence tools in the coming months in order to compare them to more traditional methods.