The Joy of Insights: How to harness qualitative data in your work

 black and white photo of a chicken and her chicks crossing a dirt road

“Data shows nine users of varying status crossed the road once at 10am from the East...why did they cross the road? Oh, we didn’t capture that.”

Do you love qualitative data? Enough to marry it? Well you should. Quantitative data gives you the hard numbers: what, how many times, when, generally who, and where. Quantitative data also leaves out the biggest and possibly most important factor: why.

I’ve come face to face with qualitative data often in my work for Library Information Technology (LIT) here at U-M. This data has been gathered both internally (such as assessment of LIT’s project intake process) and externally (surveying a variety of the University’s population about use of our “Favorites” tool). There are no hard and fast rules to employing qualitative data in your work, but I will share below the general steps I follow.

Step one: collect the data

The beginning of your whirlwind romance with qualitative data is to actually collect it. This may be a step or two outside of your normal realm of just looking at the analytics, but it doesn’t have to be hard. A few ways to collect qualitative data include:

  • Incorporating free-response text boxes on surveys

  • Meeting with your users (especially perspectives you would like to learn more from) and ask them a few questions

  • During usability tests, asking “why?”: “Why did that seem like the best place to find x?”

  • Observe your users in their natural habitat (aka work outside of your office in different library (or even campus) locations)

There are endless opportunities to collect qualitative data. You can see some in Weave UX, Putting the User First, and many, many, other great reads.  

Step two: read through your responses

It can be so very easy to just jump into the responses you get and start lumping them together. Therefore, I ask you to restrain yourself for just a little, and read through all of your responses just once: no coding, no grouping, just reading. Doing so will give you a great general sense of what your users are saying and make the next step that much easier. Anyone who has coded qualitative data will realize (at least once) that they missed a trend and will have to go back and recode earlier work. Reading through everything will hopefully help prevent doing work twice.

If you really can’t avoid the need to start organizing the results, you can jot down on the side some of the trends that recur. These trends will become your codes in the next steps.

Step three: code your data

There are so many different ways you can code data, and it is important to pick whichever you feel most comfortable with. I’ve seen complex spreadsheets, color coded word docs, and sticky notes that stretch as far as the eye can see. Pick a method that works best for your style of processing.

  • If you are a visual thinker, you may prefer sticky notes with quotations from users on them as they can be easily grouped together to demonstrate similarities.

  • If you already live in spreadsheets, you can arrange your data in different columns and weight each response (positive, neutral, negative? Vehement to apathetic?).  

  • If you have some money and don’t mind jumping into a third-party software, Dedoose is probably one of the most useful coding tools I have encountered. You can cross-reference different codes and mix together quantitative and qualitative data. You can also export your results into word documents (codes intact) in case you one day need to leave the platform.

Essentially with coding you are looking at trends, themes, and recurrences in your data. From these themes you can gain insights into what your users may be thinking and why they act in certain ways. Once you gather these insights, you get to share them out!

Step four: pick your darlings

Not all qualitative data is created equal when it comes to representing your users, and by that I mean, some responses do a better job telling the story of your prototypical user. Your collected data as a whole is something you can point to as support for your user-centered decisions, but before you get to that point you may need to advocate for change to get there. Pick a few really good quotations that best represent your users* and let these be your poster children moving forward.

As you work more and more with qualitative data you will find your army of darlings growing, your base of data strengthening, and your connection to your users improving. Eventually you will finding yourself loving qualitative data enough to marry it.

*Double check that you are representing your users, and not what you wish/want your users to be. This data shouldn’t be to fulfill your own preconceptions, but if it does anyway, consider it a bonus.


on Feb. 14, 10:25am

A nice reminder to someone who really likes look at the logs/analytics. Do you have some good ways to make sure you are representing the users, and not what you wish they were?

Katherine Marie Kosinski
on Feb. 22, 10:35am

The easiest (but also hardest) way to make sure you are representing the users is to get out and meet with them. Interview the stakeholders in a large enough sample to get an idea. Granted, this isn't feasible with the standard day-to-day workload. It is worth investing some time on in-depth user studies for persona creation, so in the future you can fall back on these, rather than recreating the user studies wheel or relying on assumptions. Does this make sense? Let me know if it doesn't!