Welcome to the Listening Post!

Surveys Suited for Text Analytics

Tweet

 

When and how should you consider using text analysis for your survey projects? That depends on the type of survey.

For small, one-off surveys, you don't need text analysis: do your survey coding of the verbatim responses manually. This will produce the highest quality analysis. This level of manual effort is needed even if you are using text analytics: think of the text analysis software as someone you are going to delegate the task to, but only after you specify it for them in detail.

For larger, one-off surveys -- say, surveys with more than 400 responses per open-ended question -- you are going want to use some text analysis tools in an ad hoc fashion to categorize responses. Doing it all manually becomes much too time consuming. Alternatively, you can take a random sample of the verbatim responses and manually code only that subset.

When you consider recurring surveys -- transactional surveys or quarterly relationship surveys -- you quickly realize that investing in automation will save you a considerable amount of labor each week, month or quarter that you run the survey. This can become attractive even for smaller volume surveys, because of the desire to create dashboards and track trends for each time period.

Does this sound you like need text analytics for almost every survey? Actually, the vast majority of surveys that organizations do are precisely the small volume, one-off surveys for which text analysis is currently overkill. Manually reading open-ended responses is here to stay.

See also:

• How Text Mining Changes Survey Analysis
• How Text Mining Changes Questionnaire Design
• Text Analysis for Transactional Surveys