Net-nets and Net Promoter Scores
February 17, 2021
Coding open-ends is a way to simplify responses by putting them into countable categories for analysis. Here are a few tips to make coding open ends easier.
Inductive or emergent coding starts with reviewing the open-end responses for recurring themes. Based on your findings, you create a list of coded categories called a code frame to classify the responses.
There is also deductive coding where you start with a predefined code frame. This code frame could be from previous research, or your expectations of which topics will be mentioned. You can quickly categorize the responses you knew to expect, and find insights in the unexpected answers.
Code all responses at once
You might be tempted to start coding responses from partial data, but this may actually cost you more time. As you receive more responses, you might need to add new codes, or edit the description of existing codes. Then you’ll need to re-evaluate how these changes affect all responses.
Consider other questions for context
Other open-ends, or closed-ended questions in your survey may provide context when you’re coding. Are the responses from a standalone open-end, or are they responses to an “other, please specify”? That will change the way you see the answers, and you might need to code some responses back into existing categories. One open-end may have similar answers to another, and you’ll want to use a consistent code frame for both.
Create a code frame before you classify responses
Whether you’re using the inductive or deductive coding method, make sure you have a code frame before you classify responses. There are two different types of code frames you can use; decide which type will work for your set of responses. Flat You can map all the responses to a set of broad codes that have the same level of specificity. This is called a flat code frame.
HierarchicalA hierarchical code frame has detailed codes in addition to broad codes. For example, “Difficult to use” is a broad code and you can have more detailed codes like “Too complex” for another level of granularity.
Cover all responses
When creating broad codes choose generic ones, so that each response will fit into one of the categories. If there are themes that are mentioned infrequently, or only once you can create an “Other” category to classify these responses.
Consolidate similar codes
Whether you have a flat code frame or you’re planning on adding detailed codes, a concise list of broad codes is necessary to reduce the data. Your coding should lessen the amount of information presented. If there are too many codes, then the result will be counterproductive. “Familiarity” and “Efficacy” would clearly be different codes, while “Familiarity” and “Personal experience” may have subtle differences but can work as a single code.
Assign multiple codes when appropriate
Some responses might fit into multiple broad and detailed codes. If you have a hierarchical code frame, some responses will be assigned multiple codes. For example “complicated and instructions are unclear” would be tagged with the broad code “Difficult to use“ as well as the detailed code “Too complex”.
Some responses might contain two ideas, for example “Product X is easy to use and affordable.”“Ease of use” and “Cost” are two different broad codes for this one response.
Create contrasting codes
Create codes that cover both the positive and negative side of the same idea. For example, the detailed codes “Difficult to use” and “Easy to use” contrast each other.
Review your coding
After each response is coded, review the results. What to look for: Good coverage: All responses should be coded, even if you must include an “other” category for certain responses. Accuracy and consistency: Sometimes it’s not realistic to read through every single response. You can skim the responses by code category and check for accuracy/consistency– if there’s a discrepancy it’ll stand out.
Code the responses twice
Different coders will have a different perspective on the responses. Ask a colleague to code the open-end and see if you get similar results.
Here we shared some tips and suggestions you can follow when coding open-end responses. You can apply this knowledge whether you’re coding in Excel, or using Protobi’s coding features: Simple recode tool Advanced recode too