GroupSolver has developed a distinct approach to collecting and analyzing open-ended responses. This approach combines traditional statistical methods in typical quantitative surveys with crowdsourcing techniques and machine learning, allowing GroupSolver to quantify natural language answers effectively. The result is a “self-building” platform that provides the researcher with validated hypotheses/answers to critical open-ended questions.
Most data analyses would not be possible without understanding the statistical relationships between answers provided by individual respondents. The most common analyses depend on understanding the relationship between variables and respondents. This is where the GroupSolver approach comes into play:
Ideation
Our process begins with recording verbatim answers to open-ended questions. Once a response is submitted, the respondent’s answer is run through a series of filters that will clean the verbatim response (gibberish, profanity, zero-value). The cleaned response enters the Idea Pool (the combined group of answers from all the respondents).
Consolidation
The Consolidation phase will occur if a respondent answers a question that is similarly worded to a previously submitted response by another respondent. They will then be asked if their response can be replaced with the other. If so, the answers will be merged, ultimately relieving time from manually working through each open-ended question.
Ideation and Consolidation Phases
Evaluation
In this phase, respondents will react to answers previously provided by others who have already taken the survey. Respondents will be asked to “agree” or “disagree” with other verbatim responses, ultimately evaluating which responses resonate with them individually.
Evaluation Phase
Synthesis
In the background, as respondent answers are beginning to filter through the platform, a dynamic and self-calibrating natural language processing algorithm organizes answers into themes. As more answers enter the platform, insights start emerging.
Understanding the relationship between answers given to us by respondents can uncover a much richer and deeper understanding of our customers. It gives us a more nuanced view of what respondents tell us rather than a one-dimensional, free-text analysis. Going from simple text analysis to quantified data sets allows us to understand the authentic customer response in a much more secure way.
To read more about the respondent experience, refer to the Respondent Interaction article of the Help Center.