Thematic Clusters: Uncovering themes among open-ended responses based on their meaning

Overview

For Ask Opinion and Ask Experience questions, Thematic Clusters strive to answer two key questions relating to your participant responses:

• What are the key themes among responses?
• And within these themes, who are the people responding?

This is an incredibly powerful and time-saving tool that allows you to quickly organize and analyze your qualitative data.

Included in this article:

1. How it Works
2. Where to find Thematic Clusters
3. Understanding your clusters
4. Using the Group Slider
5. Comparing Themes and audience Segments
6. How should I use Thematic Cluster
7. FAQs

How it Works

The Remesh algorithm analyzes each response to understand its meaning and how similar or different it is from other responses. This eliminates the process of manually identifying similar responses. You can define the number of groups you’d like to organize the responses into. Each response will map to a single group. 

As the system assigns each response into a group, it is optimizing for a best fit informed by the following ideas:

• Responses within the same group are generally similar to each other
• Responses across different groups are generally dissimilar to each other

As with any AI-driven algorithm, this process will not be perfect. There will be scenarios in which the following occurs:

• Some responses may not have an optimal fit into any group
• Some responses across different groups may in fact share some similarity (this is also affected by the number of groups set by the user)

Where to find Thematic Clusters

1. Navigate to the Analysis page and select the Results tab

2. Choose an Ask Opinion or Ask Experience question to analyze

3. Open the Group By dropdown to the right of the screen and select Thematic Clusters

4.  Expand the groups to view the individual responses

Understanding Your Clusters

You’ll see two labels that help you quickly understand the key themes inside each Cluster:

1. Common Topics: A list of prominent words or phrases that appear in responses.

2. Featured Response: The response that best represents the central idea of the group for the selected segment.

Common Topics add valuable context, but sometimes there is an overlap of Common Topics between the groups, which can be confusing. Common Topics coupled with the Featured Response provides instant clarity so you can quickly understand what’s in each Cluster.

The Featured Response will change depending on what Segment is selected. 

Note that since the Clusters are created from all responses (i.e., the “All” segment), the most representative response will also be shown when looking at the All segment.
 

Showing and hiding the Featured Response

1. Use the menu to control whether the Featured Response is visible when the Cluster accordion is collapsed or expanded. These are separate, independent controls.

2. When the Cluster accordion is expanded, showing the Featured Response means that the Featured Response will appear at the top of the response list. When the Featured Response checkbox is off in the expanded state, the response will rest in its default % Agree sort order position.

Using the Group Slider

By default, Remesh will display what is believed to be the “ideal” number of groups based on responses and population. Using the Group Slider, you can distill it down to fewer groups or expand to more groups to instantly play with the data.

For example, if you select three groups using the Group Slider, Remesh will optimize for the best fit of all responses into three clusters. If you select four groups, Remesh will optimize for the best fit of all responses into four clusters. And so on.

NOTE: Each response will appear only in one group.
 

Comparing Themes and audience Segments

• To compare themes, click Compare and select Themes from the drop-down menu.

You’ll see each Cluster, or theme, along with some key information about the Cluster:

• Theme Prevalence: The percentage of people within the population who have a response that fits within the theme.
• Response Count: The number of people within the population who have a response that fits within the theme.
• Average Sentiment: Sentiment for each response is categorized as “Positive,” “Neutral,” or “Negative.” The Average Sentiment is the average emotional direction among all the segment responses that fit within the theme.
  • NOTE: If you hover over the Average Sentiment, you’ll see a chart showing the sentiment distribution.

• Average Agreement: % Agree is an estimate of the percentage of participants that agrees with each response. The Average Agreement is the average of all % Agree scores for all responses within the theme, from the perspective of the selected segment. 
  • NOTE: If you hover over the Average Agreement, you’ll see a chart showing the agreement distribution, grouped into 10% bands.

 

You can compare segments, such as demographic segments, to surface key findings relating to your themes.

• From Compare > Themes, click the ‘+’ button in the right corner of the themes display.

• Select the Segments you want to compare and click ‘Done’
• The Segments will now appear as columns in your themes display, so you can compare the Segments to each other and to the Clusters as a whole. Learn more about what a segment is.

 

 

How to use Thematic Cluster data?

Use this data to identify similar themes based on full responses that go beyond individual words or short phrases (as performed by Common Topics). This works best when participants supply rich, thoughtful, open-ended responses.

How should I use Thematic Cluster Comparison view?

This allows you to review and understand differences, if any, between theme group responses and the demographics of the participants who supplied them.

Use this view to see how segments contribute responses towards each theme (“how many or few”). Also use this to compare spread of sentiment and agreement side by side. A simple way to do this is to 1) add related segments into view and 2) scan across the row to note differences in sentiment and agreement.

Thematic Clusters FAQs

1. What is a Cluster?
   • A Cluster represents a possible theme, or theme idea.
2. What question types can use Thematic Clusters?
   • Ask Opinion, Branch Opinion and Ask Experience questions.
3. Can responses appear in more than one Cluster/Group?
   • No, each response will only appear in one Cluster.
4. What is the grouping logic used for Thematic Clusters?
   • Thematic Clusters is powered by a clustering algorithm. Given the number of clusters/groups (either the default or as set by the user), the system assigns each response into a group. It is optimizing for a best fit informed by the following ideas:
     • Responses within the same group are generally similar to each other
     • Responses across different groups are generally dissimilar to each other
5. How is the Thematic Cluster Comparison view organized?
   • Summarized information about the responses within each cluster/group are shown by segment, and includes details like segment size, sentiment, and agreement (for Ask Opinions and Branch Opinions).
6. What is Theme Prevalence?
   • The percentage of a segment represented within a specific theme.
7. How is the Average Agreement score calculated?
   • It looks at the responses within a theme and averages out the agreement for all the responses from the perspective of that segment.
8. How is the Average Sentiment label calculated?
   • It looks at the responses within a theme for a specific segment and averages out the sentiment for those responses.
9. What languages is Thematic Cluster available for?
   • German
   • English
   • Spanish
   • French
   • Italian
   • Japanese
   • Dutch
   • Portuguese
   • Chinese (Simplified)
   • Chinese (Traditional)
   • Arabic
   • Czech
   • Danish
   • Finnish
   • Hindi
   • Hungarian
   • Indonesian
   • Korean
   • Malay
   • Norwegian
   • Polish
   • Romanian
   • Russian
   • Slovakian