Methods

Introduction: Coding and Analysis

Coding is the process of looking through raw qualitative data to discover. In his book, The Coding Manual for Qualitative Researchers, Johnny Saldaña posits that coding helps confirm “our descriptions of people’s “five Rs”: routines, rituals, rules, roles, and relationships.”

Besides defining a research question, coding and analysis is the step when design researchers have the most agency in the research process.

Simply, the process of coding and analyzing data involves two steps. First, researchers look through data and tag content with “labels” (codes) that make the data easier to grasp and group. After this, these codes are compared to find similarities, trends, and patterns. To start this process, data has to be accessible (printed, posted, visualized) so bits and pieces can be thoroughly and consistently labeled.

Tagging Instances (Coding)

If you had data in the form of photos and interviews, every time a positive emotion would appear in the data, it could be tagged as “happy,” and every time a negative one would happen, a tag of “sad” would apply.

“I hate that my handbag gets sticky when I set it on the ground in a movie theater” (sad tag)

A photo that shows a frustrated emotional face when a person sits in a seat at a movie theater (sad tag)

This is a very basic example (because people are seldom just “sad” or “happy”), but it gets to the heart of coding data.

Themes are the products of this process: “a-ha!” moments in research when patterns or trends coagulate to form an unarticulated area of need, a deeply held belief, or a pattern of behavior that may not have been noticed before.

Coding for Design Research

For design research, the goal of coding data is to find themes that inform design: what to make, how to make it, where to deploy it, how it should work, etc. Your ability to find themes based on evidence will equate to your ability to innovate. It’s all hidden in the data. Coding is what makes the intangible, tangible.

Hunch-Level Coding

Sometimes all we need is hunch-level evidence to support design decisions. In these cases, coding can be pretty basic, and a small range of coding levels will work. In the video below, I demonstrate hunch-level coding using a range of five smiley faces.

Detailed Coding: Visual Data Examples

Coding with a range of happy faces fails to capture all the details of rich experiences. Sometimes, you have to make your codes based on the data. In the video below, I demonstrate how to code visual materials like photographs by creating codes on-the-fly that emerge from the data.

Coding Data for Discovery

I am leading a research group that is exploring how we can design ways for people to record their end of life choices. I shot the following video during the coding and analysis process. Here’s your chance to see the detailed coding process in real-time.

Coding Wisdom

“When we walk with the wise, we will be wise.” This statement is a good reminder that others who have gone before us can show us the way because they have experience. Here are a few tidbits of wisdom on coding based on my experiences.

Use the same codes across all of your content: when you do your first pass over the data, make up codes for any content that sticks out. Re-use codes when you see repeated content instances later in the data. Over the course of a coding process, you’ll amass a healthy list of codes that can be applied to all the data, regardless of the media.
Do multiple passes: Create codes for content as you go when you do the first pass through the data. Once you’re done, go back through all your data again and apply codes from your growing list. When you do this second pass, you’ll likely see where codes you created later in the process can be applied to the first bits of data you reviewed in the first pass.
If you feel like you’re flying blind, that’s normal: The first time through the data, you may have a hard time defining codes. Do your best to label whatever makes sense. When you do a second pass, things will become a lot clearer for what codes make sense and what codes weren’t effective.
Code individually, then come together as a team and compare: We all see things differently, and it’s one of the best assets of working as a research team when it comes to coding data. Not everyone will see the same things in the data and not everyone will use the same codes. When this process is completed individually then codes are compared, overlaps between different peoples’ coding will confirm the validity of those codes. For the codes that don’t overlap, work together to figure out if these need to be coordinated or discarded.
Create a WTF code and a great quote code: I learned these tips from Jennifer Heston, a researcher with Scripps Gerontology Center at Miami University. The WTF code will help you group all those statements that don’t quite fit anywhere. Sometimes, groundbreaking discoveries can come from surprise statements. The great quote code will help you find statements that you may want to build into a presentation or report later on. If you code these on-the-fly, they will be a lot easier to find later on.

Tools for Coding Data

If you’ve done your work correctly, you probably have a ton of data to parse through. That means you are now sitting on a pile of dozens of videos, hundreds of photos, and thousands of words. There are lots of tools you can use to code this data. Just make sure your research team all uses the same tools for ease of coordination.

Digital Tools that Do It All

These tools are not free, but they allow researchers to upload all types of media and code on the fly. They’ll also count up all of the code instances and spit out fancy reports for you.

Dedoose: web-based
ATLAS.ti: software
SPSS: software (for quantitative data)

Coding Tools to Mix and Match

Projects in this course don’t involve massive data sets, so mixing and matching tools can work just fine.

Microsoft Word
Adobe Acrobat
Adobe InDesign
Paper and markers (yes, hand marking content works very well!)