It is an alchemist’s blend of art and science when it comes to qualitative marketing research. The analysis of open-ended questions found on most surveys is indeed a qualitative endeavor. At its core this is an iterative process, even with the use of text analytics tools such as OdinText.
Where do you start? In my experience a good place to start with text analysis is with a word cloud. These charts provide a visual cue of the importance of words and phrases. The larger the typeface, the more relevant the word is to the data. By relevance I mean number of mentions. The tools on the market now will create a ranking based upon the raw number of mentions and the associated percentage. For an example see the picture below.
![word cloud]()
The word cloud provides you a starting point for creating categories. The goal here is to categorize as many responses as possible, with the understanding that some comments may not be relevant and best left uncategorized. These categories should be viewed as multiple response options, in the sense that a respondent’s passage can have any or all of the categories you have created.
There are two primary considerations when analyzing text data, with category development being the first. Tone is the second concern. This is an area where the researcher needs to bring his or her expertise to the table. Assessing the tonality of a passage is a modestly difficult task for an automated tool. In the context of a customer satisfaction survey the researcher would need to evaluate the comments to assess whether or not they were negative, positive or neutral. This is also known as sentiment analysis.
Traditionally there has been a high degree of subjectivity involved with qualitative analysis. More advanced tools, like the aforementioned OdinText, can remove much of the subjectivity. However, in absence of an automated tool, it is recommended to have another set of eyes review the categories you have created.
Open-ended comments, be it from surveys, online chat-logs or other sources can provide a wealth of untapped consumer insight. The raw data can be imported from existing studies or collected as part of a current research stream. Either way, text analysis can facilitate a deeper connection with your respondents.
The word cloud provides you a starting point for creating categories. The goal here is to categorize as many responses as possible, with the understanding that some comments may not be relevant and best left uncategorized. These categories should be viewed as multiple response options, in the sense that a respondent’s passage can have any or all of the categories you have created.
There are two primary considerations when analyzing text data, with category development being the first. Tone is the second concern. This is an area where the researcher needs to bring his or her expertise to the table. Assessing the tonality of a passage is a modestly difficult task for an automated tool. In the context of a customer satisfaction survey the researcher would need to evaluate the comments to assess whether or not they were negative, positive or neutral. This is also known as sentiment analysis.
Traditionally there has been a high degree of subjectivity involved with qualitative analysis. More advanced tools, like the aforementioned OdinText, can remove much of the subjectivity. However, in absence of an automated tool, it is recommended to have another set of eyes review the categories you have created.
Open-ended comments, be it from surveys, online chat-logs or other sources can provide a wealth of untapped consumer insight. The raw data can be imported from existing studies or collected as part of a current research stream. Either way, text analysis can facilitate a deeper connection with your respondents.
