Surveys are often completed by thousands of people, all with vastly differing opinions. You'd think it'd be impossible to quantify all of them right? That's where coding survey responses comes in and it's why the rest of the GreatBlue staff likes our data analysts so much.
To understand how coding a survey response works, we first have to discuss the two main types of survey questions:
- Close-ended questions have a limited number of predetermined responses (think yes/no or agree/disagree) and are easier to compile into a single dataset as each predetermined response is assigned a number used for coding ahead of time. For example, if it's determined no = 0 and yes = 1 when designing the survey, then it's much less labor intensive to perform certain statistical analyses since we don't have to manually assign every single question response to a number after the data has been collected.
- Open-ended questions on the other hand are much more complex and are the reason we appreciate our data analysts so much. Open-ended questions, as the name implies, can have an infinite number of unique responses and are much trickier to code.
Open-ended questions are where data analysts make their money. Like we mentioned above, open-ended responses can have any number of unique responses. And, depending on the topic, some people have a lot to say.
The easiest and most common way for open-ended responses to be coded is by bucketing them out. What this means is the data analyst will compile every single open-ended response and look for similarities between them. Responses with similar themes are placed in the same bucket, and quantified (just like no and yes were in the close-ended example above). After this is accomplished, the analyst can proceed with statistical analysis.