We are using our own LLM to categorize responses. The categorization happens in three levels (drill-down principle):

• Category 1 (e.g., Product)
• Category 2 (e.g., Ease of Use)
• Category 3 (e.g., App navigation)

Each response can have multiple categories. We are using an API-based approach to determine the categories after a response is captured, and then we want to write them back to the Qualtrics response dataset.

The ultimate goal is to visualize these categories in a dashboard – for example, with bar charts. However, we want to implement a dependency: if Category 1 = Product is selected, the rest of the dashboard should automatically filter to that subset. Subcategories that do not belong to the selected Category 1 should no longer appear.

Our challenge: How can we achieve such a visualization? In which format should we write the categories back to the response dataset?

Initially, I thought about multi-value fields and codes, e.g., values like „1, 2, 5“ (Category 1) or „1-2, 2-2, 2-6, 5-1“ (Category 2) and so on. However, honestly no idea how to achieve the dependecy and how to implement a code list for that…