The ABCs of Coding Qualitative Data for Beginners

Editor's Note: Take a look at our featured best practice, Turn a Business Problem into a Data Science Solution (15-page PDF document). Learn about the most common machine learning algorithms in regression, classification, and clustering problems which include linear regression, logistic regression, naive Bayes, support vector machine, and k-means. Then an introduction to the Python code for a classic classification problem, [read more]

* * * *

When conducting qualitative research, surveys and interviews can generate the most valuable information. Free-form texts or open-ended questions enable participants to give more detailed feedback.

Yet, analyzing these data can be challenging. This is because they don’t have numerical statistics to back them up. They are not encapsulated in complex formulas to ensure definite results.

This is where coding comes in. It can help you transform messy scripts or responses into quantitative metrics. It allows you to analyze the data, observe recurring patterns, and create a data structure. With it comes an assurance that your research is accurate and free from biases.

Coding Qualitative Data

In a nutshell, coding allows you to assign quantitative tags to every piece of data. Codes can be words, phrases, and sentences. You can even use paragraphs as codes if you want to. As long as they fit your data, they will work and help you with an accurate interpretation.

It is pivotal in a large-scale analysis since it ensures credibility. You need to be consistent in comparing and contrasting each piece of information. With coding, you can use spreadsheets to quantify and classify the data.

For example, a customer writes a hotel review on TripAdvisor. “The hotel provides great service, but the room was pretty expensive.” You can assign codes as tags like the quality of service and price.

There are several ways of coding qualitative data. You can do it through manual coding or by software. But before gathering data, you have to choose between deductive and inductive coding.

Deductive Coding

Deductive coding is best for the first pass or initial coding. Here, you need to create predetermined sets of tags or codes before the actual process. These may vary in line with your research questions and framework. In short, codes may vary as your analysis goes on.

For example, you want to conduct online surveys or reviews on hotels. You may use the price, quality of service, and name or brand as your predetermined tags or codes.

If you wish to get feedback on a specific restaurant, you may use price, quality of food and service, and its name. Note that you can’t use methods like Likert Scale since you are working on free-form responses.

Inductive Coding 

Inductive coding is best used if you don’t have preconceived ideas about a topic yet. You will have to create from scratch or develop codes based on the data gathered.

With this, you will have to go through it and familiarize yourself with creating codes. It can be lengthy since you’ll have to understand it first before assigning codes. But, it can avoid biases compared to deductive coding.

Three Steps for Coding 

When you are coding qualitative data (like this), you have to follow the steps below. Let’s assume that you wish to get customer feedback on a specific restaurant.

1. Start with Broad Categories and Narrow Them Down

In initial coding or first pass, you only have to get familiar with the data. You may start the process by assigning codes to broad sections or categories.

Even if you don’t intend to take everything into your final narrative, it will help you understand it better. You can see some patterns and recurring ideas.

After doing so, you have to have a keen eye when checking every broad code or tag. You may remove redundant or less necessary codes to have a more precise analysis.

Then, you can sort them and group them into similar categories or subcategories. From here, patterns and ideas will become clearer.

2. Assign Emotions or Sentiments to Categories 

From each category, you may assign sentiments or emotions. For example, you used the price of the food, the quality of service, and the restaurant’s ambiance. You can group them into positive or negative emotions. You can move them around to see a structure that suits your analysis.

As you can see here, it appears that you’re doing inductive coding. Note that you can use both deductive and inductive coding in your data. Given this, you can continue to narrow down categories. You can now figure out your scale and measurements.

3. Take These Categories and Sentiments into Your Final Narrative

Once they can reflect recurring themes, you can start quantifying them. You can now understand the stories behind the data. You are free to create a meaningful conclusion.

For example, you have surveyed 100 people. Seventy of them fall under the food-positive sentiment. Meanwhile, 20 of them fall into service quality-negative sentiment. The remaining 10 are under the restaurant ambiance-positive sentiment.

This will show you that the restaurant serves good food at a reasonable price. But, there’s still a need for customer service improvements.

Four Tips for Accurate Coding 

1. Start with a Small Sample of Data 

It does not mean that you have to compute the sample size from the population before doing the survey. This process is straightforward. You only have to get a specific percentage from the data after categorizing the codes.

From the example above, you may get 10% of responses from each category and analyze them. If you can see recurring themes and come up with a possible answer, you can proceed to the remaining data.

2. Use Numerical Scales

You may do this after assigning positive or negative sentiments to every category. You may wish to know how positive or negative the sentiments are on a scale of 1-5.

For example, if a customer says the food is very delicious, you can use 5/5. If a customer says crews are quite unattentive, you may use 3/5 to quantify the service quality.

3. Keep in Mind That Each Data Can Have More Information

This is why you have to read through your data again after doing the initial coding. Check every code twice or thrice, especially if they can be also placed in other categories. Doing this is important to maintain consistency and lessen biases.

4. Be Mindful of Too Many Codes

It’s okay to assign broader codes to different data. But, do not assign too many codes so you won’t have a hard time sorting and analyzing them.

In the same example above, some customers may comment on cashiers. Others may comment on servers and security guards. You can put them in a single category like service quality to simplify it.

Qualitative research appears easier without its complex formulas. Yet, it can be more challenging since responses are often unquantifiable. But, coding makes it possible while reducing biases. It can be tedious and time-consuming, but it guarantees credible outcomes.

Do You Want to Implement Business Best Practices?

You can download in-depth presentations on Data Analytics and 100s of management topics from the FlevyPro Library. FlevyPro is trusted and utilized by 1000s of management consultants and corporate executives.

For even more best practices available on Flevy, have a look at our top 100 lists:

• Top 100 in Strategy & Transformation
• Top 100 in Digital Transformation
• Top 100 in Operational Excellence
• Top 100 in Organization & Change
• Top 100 Management Consulting Frameworks

These best practices are of the same as those leveraged by top-tier management consulting firms, like McKinsey, BCG, Bain, and Accenture. Improve the growth and efficiency of your organization by utilizing these best practice frameworks, templates, and tools. Most were developed by seasoned executives and consultants with over 20+ years of experience.