deepsight cloud · Topic analysis Organize open-ended responses by topic.
Discover new categories or work with your existing codebook – and check the assignments against the original responses.
One topic, different wording
Responses worded differently end up in the same topic. A single response can be assigned to several topics.
“The hotline is always busy.”
“I just can’t get through by phone, email worked in the end.”
“The new price plan overview is easy to understand, but I couldn’t find the cancel button.”
Select a topic to see the responses behind it.
Illustrative example. “Cancellation” was added during the review and saved as a new topic version.
New topics or your codebook
Afterwards, you can rename, merge, split or remove topics – each change saved as a new topic version: Topic refinement.
Need a custom-trained model? We assess this as part of a custom project: Custom-trained models.
Questions about topic analysis
You decide up front how detailed the topics should be; this sets the approximate number of topics. deepsight may also find fewer. Anything that doesn’t fit a topic ends up in a residual “other” category.
A keyword search only finds the terms you search for. Topic analysis groups responses by meaning: “too expensive” and “not worth the money” can end up in the same topic, even without a shared word.
Yes. In the study context, you add the research question, objective, target group and key terms of your study, such as product or brand names. Topic analysis draws on this context.
Shares and figures refer to the analyzed responses; anything the data quality check filters out does not count towards the base. Because a response can be assigned to several topics, it counts towards each of them.
You upload it as an Excel (.xlsx) or CSV file, optionally with a short description and the parent category for each category; both are used for the assignment. You can save newly discovered and reviewed topics as a codebook for the next wave.
Working with a given codebook is common for market research agencies handling client projects.
More features · Analyze responses
See the feature with your own data
We will walk you through the analysis using an example from your field.