Cracking the Qualitative Research Code

Your inferences are logical, and your insights inspirational. Even your PowerPoints are sexy. Audiences weep over your video, and yet, somehow, it’s not enough. Numbers-obsessed executives dismiss your work. What gives?

The missing puzzle piece takes shape from a few thoughtfully acquired, well-placed digits to amplify the compelling words of your respondents.

Some of us have learned how to soothe the most scrutinizing figure-fanatics, persuade the c-suite to embrace qual insights, and even to evangelize the narratives. One of dscout’s goals for this new publication is to empower our research compatriots with the knowledge and tools to do the same.

One critical storytelling gizmo is one you already know a bit about: coding.

Keywords are for blogs, not research

Researchers know coding as a process of compiling, labeling, organizing and sorting data. For most, it happens early, as a base for analysis. For some, coding is analysis. For others, codes simply note what piques their curiosity or seems interesting.

At dscout, coding is key to communicating to clients. We code to synthesize our qualitative findings, and also to add numerical credibility and intrigue to our insights.

Where would you rather be?

Boardroom A

As you relay an enlightening anecdote from Respondent #3 to support your insight, Vice President Eeyore yawns, then squints at your PowerPoint slide. “I don’t give a crap what one guy from Tallahassee has to say. Let’s move on.”

Boardroom B

Just before clicking play on your respondent video montage, you reveal that 252 responses independently referred to your top insight. VP Eeyore looks up, stays quiet, leans forward.

Join us in Boardroom B! The entrance is in the coding room.

Code comprehensively, and oodles of data can emerge to support your insights. Approach it superficially — coding just the eloquent responses or with only descriptive meta tags — and you’ve got yourself a nifty little set of keywords, but not much else.

Keywords are great for a blog. You can do more for your research story.

The best coding process begins once your story starts to emerge, and you’ve identified the dimensions you care about. Until then, you can waste a lot of time and effort and make a mess.

Deciphering the Process

Many researchers start coding right away in an attempt to draw out the story. Since we approach it differently, we asked dscout insights lead Carey Palmer to talk about how, when and why to code:

1. Hands-off immersion

I don’t start coding right away. At first, I immerse myself in the responses and generate a list of things I think are the start of insights or themes. Absorb the responses and think about my end game.

Say a client’s looking for a relationship between food and ‘family fun.’ Maybe you’ll end up creating personas that link the two. Eventually you’ll code for that.

2. Frameworks and stories

For every project, we establish a conceptual framework — journey mapping, for example — as a foundation for the story. Start coding once you have a story that you want to validate or quantify some aspect of.

When I think I have a good storyline going, and a framework that simplifies what I’ve learned, and I understand what the interesting bits are, then I code according to the framework.

3. Start with a sample…

Coding is an iterative process, a living one. Your system and your data start to interact. Patterns shift as you start applying codes.

We start by coding a small sample to confirm our inventory of codes is correct and complete. You really do not want to do it in multiple rounds. Touching 600 responses twice will really slow you down.

4. …but code every single one.

Don’t quit coding after you’ve pegged the insights. Tag every single response according to your framework.

Being thorough is what transforms your insights from anecdotal evidence (“Respondent #9 said it this way”) to the quantifiable gold (“Half of all respondents mentioned this!”) that gets the nod of decision-makers and supports your narrative.

Of course, thousands of responses requires a lot of coding. It means blocking out the hours to just get it done.

I kind of think about it like working out. It helps to do it with someone. It’s like a marathon.

OK, maybe it’s more like a lock-in.

5. Code for color

Easy-bake, descriptive meta tags are usually the codes you think of first, but those are most useful after you’ve identified and coded for your framework.

We might start with a list that’s descriptive and straightforward, such as “kids, mom, kids plus mom.” Those aren’t about the story, just a way to segment data. If we also have a set of “behavior modes,” (like “outdoor play,” “couch potato,” etc.), we apply a mode to every response. Then we can compare modes by using the descriptive codes.

I think that’s what’s really cool about coding. And, that’s when people are really impressed, like, “46% of responses were about this? Oh, wow, that’s really something to look at.”

It never hurts to export your data to a spreadsheet to compare codes.

(Search YouTube for a 5-minute tutorial on Cross-tabs, aka Cross Tabulation, aka Pivot Tables, aka Contingency Tables).

Serendipitous discoveries like Carey found on that project are wonderful, and they are rare. So we don’t depend on coding to create our narrative, only to support it.

Applying codes comprehensively helps internalize the research, forcing you to look more analytically at what you’re seeing. And that’s half the value.

Becoming really familiar with your data and feeling like you’re an expert on the attitudes and behaviors you saw will make you a research star. Add a few strong numbers to your insights, and you can become a major force in the boardroom, too.