Please note that some of the content below related to Analysis Perspectives is related to our upcoming November 11th release! If your experience in the platform does not reflect what is outlined in that section, please return on or after November 11th.
Overview
Quickly and effortlessly generate initial summaries from your Remesh Conversation. Summarize is a new feature that leverages a large language model (LLM) from OpenAI, the creators of ChatGPT.
LLM technology is used to power a few of Remesh's most powerful features, and does require an extra step to enable. Read more about this suite of features and how to enable them here.
Included in this Article
- How it works
- Where to find Summarize
- Conversation Summary
- Question Summaries
- Question Drilldown and Analysis Perspectives
- Analysis Perspectives Example
- Customize Summarizations
- Frequently Asked Questions
How it Works
First, Remesh Summarize generates a summary for each Poll, Rank and open-end question in your Conversation. For open end or Ask questions the system selects key responses to summarize from each question that optimize for the spread of thought, overall performance, and technical limitations around LLMs. Our Summary process includes our existing proprietary analysis algorithms in conjunction with Large Language Model prompts to provide high quality and targeted summaries. For closed end Poll and Rank questions, the summaries reflect the aggregated data collected.
Next, Remesh Summarize creates a comprehensive Conversation-level summary based on the individual question summaries. Users are able to create new slices of Conversation-level summaries by selecting their own set of questions.
Lastly, Summarize provides analysis perspectives which takes these summaries and identifies collective resonance around common themes and where key differences in themes arise across demographics.
Where to find Summarize
- After logging into Remesh, navigate to the Conversation you want to analyze
- Select Summarize from the top navigation
- You will first see onboarding polls (if applicable), then your discussion guide questions organized by section below.
Conversation Summary
What is it? Users are able to review a Conversation-level summary on the right side of the screen. This summary will contain a combination of all questions that have been selected using the check boxes on the on the left hand side of their screen.
How can I use it? Use this summary to get an understanding of key take aways from your Conversation. You can also copy and paste this summary and share it with key stakeholders after your Conversation so they can better understand initial insights.
Question Summaries
What is it? Users can review question-level summaries for each question on the left side of the screen. These are summaries of just the responses for that specific question.
How can I use it? Use these summaries to identify questions you’d like to investigate further.
Question Drilldown and Analysis Perspectives
What is it? By clicking on the blue arrow beside each question, you are able to view what data is fueling the summary and additional insights found in Analysis Perspectives. The following are part of the Analysis Perspectives insights.
- Consensus: A summary of ideas with high consensus across demographic segments. Includes 3-5 ideas from the question, with high consensus across the demographics defined by onboarding polls. Available in Ask Opinion, Branch, Poll and Rank questions.
- Divergence: A summary of ideas where there is divergence across demographic segments. Includes 2-3 ideas from the question, where there is divergence across the demographics defined by onboarding polls. Available in Ask Opinion, Branch, Poll and Rank questions.
- Themes: Top ideas from the question, presented in bullet point form, with percentage of responses that align to the idea, and a key quote that aligns to the idea. Powered by Auto Code technology. Key quotes are indexed on Agreement where available. Available in Ask Opinion, Branch and Ask Experience questions.
- Insights: Synthesized findings in bullet point format. Available in Poll and Rank questions.
- Demographic Breakdown: A comparison of demographic segments including similarities and differences that surfaced in paragraph format. Available in Poll and Rank questions.
Analysis Perspectives, Feature by Question
Consensus | Divergence | Themes | Insights | Demographic Breakdown | |
---|---|---|---|---|---|
Poll | Yes | Yes | No | Yes | Yes (will be removed later in year) |
Rank | Yes | Yes | No | Yes | Yes (will be removed later in year) |
Opinion Ask | Yes | Yes | Yes | No (coming later in year) | No |
Branch Ask | Yes | Yes | Yes | No (coming later in year) | No |
Experience Ask | No | No | Yes, shown for each category defined | No (coming later in year) | No |
How can I use it? Use Analysis Perspectives to quickly identify decision-ready insights around demographics within quantitative and qualitative data. You could copy insights directly from Analysis Perspectives to use as the basis for reporting, then build visuals in Results to further support.
Analysis Perspectives Example
Below you can find examples of what you might find in analysis perspectives for an open end question.
Market Research | Employee Research | |
Consensus | Analysis of this question revealed widespread agreement on the importance of clean ingredients in child friendly snack products. This was evident across brand shoppers and competitor shoppers alike. | Analysis of this question showed a high level of agreement on the need for better mental healthcare services offered through the employee wellness program. Requests for in-person care was a common theme along with more flexible options. |
Divergence | Analysis of this question showed a difference in importance of price for snack products amongst segments like household income. Those who indicated Below Median Income showed higher agreement that price was an important purchase factor, whereas those who indicated Higher than Median Income showed less agreement on price importance. | Analysis of this question indicated divergent opinions among segments, like employee level. Participants in the Individual Contributor group demonstrated higher agreement on the need for more flexible, remote work options than the Executive Leadership group. |
Themes | • Health Factors, Ingredients (68%): “Ultimately, we choose snacks that are safe and healthy for our family. We don’t want to risk anything with young kids in the house.” • Price Point (39%): “It seems like we’re still feeling the effects of inflation when we buy groceries, so we have to mindful of how much snacks cost to fit to our budget.” • Package design (21%): “Our family loves trying new things so it’s helpful when food companies can make their boxes standout on the shelves at the store. We’ll usually look at the new snacks that catch our attention and decide if it’s worth it.” |
• Mental Health (56%): “More flexible mental healthcare options like in-person therapy coverage is definitely something missing from our wellness plan.” • Work Location Flexibility (42%): “Previously having the option to work remotely was a huge plus. It was probably the best part of the benefits and wellness at this company. Can we bring that back?” |
Customize Summaries
You can customize your question-level and Conversation-level summaries in multiple ways. These features can be used independently or layered with each other to generate more specific summaries for targeted groups.
Segments
Segments are created based upon Polls in the Conversation, and are a way to cut the data as you analyze. Whether you’re generating summaries at the question-level or Conversation-level, you can find the Segment filter dropdown menu on the top-level page.
When filtering by segment, there may be time delays. The blue checkmark icon indicates that the data should populate quickly. Segments with a “~60s” clock icon have data that may require some loading time. Upon loading, you will see that the question-level and Conversation-level summaries will be adjusted to reflect that specific segment.
Question Selections
You can customize your Conversation-level Summary by selecting the specific questions you would like to be included, and clicking the regenerate button.
Conversation-Level Summary Type
You can adjust the type of Conversation-level summary you would like to generate by selecting the drop down above the Conversation-level summary.
-
Paragraph: The default format, around a single paragraph
-
Half-Page Summary: A few paragraphs
-
Bullet Point Summary: Individual bullet point instead of paragraph
Frequently Asked Questions
- Why can’t I access the Summarization feature?
In order to access this feature a Workspace Admin must first enable Generative AI features. Read more about that here. - I selected different questions to include in the summary, why isn’t the summary text changing?
In order to generate a new summary, you must both select the questions you want included and click the regenerate button 🔁. - How do I copy the summary text to paste into my presentations?
Simply click on the ‘Click to Copy’ button to copy the summary text to your clipboard. - Does this work on previously run Conversations?
Yes, you can use Summarize for previously run Conversations. Please note that there will be some initial loading time the first time you access your past Conversation. - Why do the % Agree values listed in Consensus sometimes differ from those in the response list?
Consensus is determined by finding the bridging (or minimum) agreement across all autosegments considered. This means that % Agree values reported in Consensus may differ from those in the Summarize list view (this is the % Agree for All participants), though they can be found in Compare: Agreement in Results. These bridging values are often accompanied by the language "at least... % Agree" or "no less than... % Agree".