Open End Data Analysis & Theme Detection
Theme Detection helps you make sense of your open-end data. Instead of reading through thousands of responses one by one, Lisa automatically identifies the key themes across your conversations. You can generate themes from your data, edit them to fit your analysis needs, customize how they're grouped, and explore what your respondents are really talking about, all from within the Visualizer.
Written By Efe Kiremitci
Last updated About 1 month ago
Identify Questions with Theme Detection
Not every question in your conversation has theme detection available, it needs to be turned on per question. In the Builder, head to Dialogue Settings (add link here) and enable Theme Detection for each open-end question you want to analyzed. Once it’s on, Lisa’s open-end pipeline will start working for that question and starts processing responses for that question.
To find which questions are ready for theme detection, head to the Visualizer. On the left panel variable view, look for the octopus icon next to a question, that’s your signal that theme detection is available and ready to go.
The octopus icon has three states:
Forgot to turn it on while building your conversation? No worries, you can enable it at any point. Just keep it in mind that Lisa will need some time to go back and process all the data that’s already been collected.
⚠️ Important: Once themes are assigned, they're final. You won't be able to re-assign or change them, so make sure you're happy with your themes before moving to the assigned state.
P.S. Curious why an octopus? Check out the open-end visualization and you’ll get the idea 😉
Generate and Assign Themes
1. Estimate Themes
Click on the variable with the grey octopus icon to open the theme detection view.
From here, click Estimate Themes. Lisa will use the context of your project and the question to analyze the open open-end responses and start identifying the key themes of your respondents are talking about. This may take a moment depending on how much data has been collected, and you can track the progress while Lisa is estimating themes.
2. Review and Customize Themes
Once Lisa finishes estimating, the octopus chart (a Sankey chart) appears, this is where you get your first look at what respondents are talking about.
By default, Lisa generates up to 7 themes with up to 5 sub-themes each. But these are a starting point, not set in stone.
To start customizing, click on the pencil icon. Click the dropdown arrow next to any theme name to expand it, this reveals the description box and the sub-themes underneath that theme.
This opens a modal, and from here you can:
Edit theme and sub-theme names to better match your analysis framework
Add a description for each theme to help Lisa estimate more accurately (think it as prompting an AI chatbot to shape your output), the more context your give about how you want to classify the responses, the more focused the results will be.
Customize how themes and sub-themes are group and structured
Estimate, once you’re happy with your updates, click estimate to re-estimate your themes based on your input.
Not happy with how themes turned out? Click the refresh icon to overwrite all your edits and start from scratch.
You can estimate themes as many times as you want while fieldwork is live, even as new responses are coming in. Lisa will use all processed responses up to 1,000 respondents. Once your sample goes above 1,000, Lisa randomly selects 1,000 respondents to estimate from.
Why 1,000? Our product team found this to be the sweet spot, it’s accurate enough to confidently roll out themes to the rest of your sample.
3. Assign Themes
Once you’re happy with your themes and sub-themes it’s time to lock them in. Click Assign Themes, and then Confirm and Assign. After this, Lisa will start assigning the themes you’ve defined to the rest of your sample.
⚠️ Reminder: This is a one-way action. Once themes are assigned, you won’t be able to change or re-assign them. Make sure you’re confident in your themes before confirming.
Once the assignment is complete, the octopus icon next to your variable name will turn green, that’s how you know everything is set and your data is fully categorized.
