Skip to main content
Topics are the thematic codes Pulse Classifier assigns to each response. Together they make up your code plan — the controlled vocabulary the AI uses when reading responses. The quality of your topics directly determines the quality of your classification results, so spending a few minutes reviewing and refining the AI-generated plan pays dividends in accuracy.

​
Auto-generating topics with AI

The fastest way to build a code plan is to let the AI read a sample of your responses and propose topics automatically.
1

Open the Topics editor

After uploading your data and configuring your columns, the Topics editor opens. If you have not yet generated topics, you will see the Generate Topics button.
2

(Optional) Add guidelines

Enter any instructions in the Guidelines field — for example, “focus on product feedback only” or “merge any topics with fewer than 5 responses”. The AI uses these to shape the proposed plan.
3

Click Generate Topics

Click Generate Topics. The AI reads up to 400 response samples and proposes a topic list with estimated occurrence counts for each topic.
4

Review and refine

Review the proposed topics. Rename, reorder, add, or delete topics as needed before proceeding to classification.
If you are not satisfied with the initial proposal, update the Guidelines field and click Generate Topics again. The AI incorporates your feedback and produces a revised plan while keeping the previous results as context.

​
The Topics editor

The Topics editor displays your code plan as an editable list. Each row shows a topic label and, once counts are available, an estimated occurrence badge.

​
Adding topics

Click the + icon on any row to insert a new topic below it, or click Add topic at the bottom of the list to append a new row. Type the topic label directly in the text field.

​
Renaming topics

Click into any topic label field and type to rename it. Changes are saved automatically as you type.

​
Reordering topics

Drag any topic by the grip handle on the left to reorder it within the list. L2 sub-topics can be dragged within their parent group.

​
Deleting topics

Hover over a topic row and click the trash icon that appears on the right to delete it.

​
Two-level topic hierarchies

Pulse Classifier supports two-level hierarchies: L1 parent topics and L2 sub-topics beneath them. Use hierarchical topics when a broad theme is too coarse on its own. Example: 
Price (L1)
  ├── Too expensive (L2)
  ├── Good value (L2)
  └── Hidden fees (L2)

​
Promoting and demoting topics

  • Click the → (indent) arrow on a topic row to make it a sub-topic of the topic above it.
  • Click the ← (outdent) arrow to promote a sub-topic back to the L1 level.

​
Auto-generating sub-topics

Hover over any L1 topic and click the sparkle icon. The AI generates suggested L2 sub-topics for that parent, scoped to the responses that match it.

​
Importing topics from Excel

If you already have a code plan in an Excel file — for example, from a previous wave of the same survey — you can import it directly. The import expects a sheet named Topics or Codes with at minimum a label column (Libellé, Label, or Topic). An optional Level column (L1 / L2) and Parent column control the hierarchy.
When you import topics from a file that also contains a Topics sheet, Pulse Classifier automatically detects the sheet and reads the topic count from it.

​
Estimated counts

After generating topics, Pulse Classifier displays an estimated occurrence count badge next to each topic. These counts are approximate estimates based on a keyword scan of your responses — they help you identify topics that may be too narrow (very low count) or should be split (very high count). Click a count badge to navigate to the codes view filtered by that topic after classification is complete.

​
Tips for good topic design

  • Aim for 8–15 topics at the L1 level. Fewer than 8 topics often loses important nuance; more than 20 can make classification noisy.
  • Make topics mutually exclusive where possible. If two topics frequently overlap, consider merging them or making one an L2 sub-topic of the other.
  • Use clear, unambiguous labels. Short phrases of 1–5 words work best. Avoid jargon that the AI may not associate with your responses.
  • Add an “Other” topic as the last item to catch responses that do not fit any defined topic. Pulse Classifier adds this automatically unless you remove it.
Changing topic labels after you have already run a classification means you will need to re-classify to get accurate results. Renaming topics does not retroactively update previously classified responses.