Editing and Refining Categories in Text Analytics¶

A generated category set is meant to be a fast first draft - close enough to save you the blank-page problem of coding open-ends from scratch, but not something to classify against blindly. This is where you make it actually match how you think about the data, before it gets applied to every response.

Reviewing the Generated Set¶

Before classifying anything, read through the drafted categories the same way you'd review a colleague's first pass at a codebook. Look for categories that are too broad to be useful, ones that overlap so much they'll be hard to distinguish once applied, and any real theme in the responses that didn't get its own category at all.

Renaming¶

A category's name is fully editable - if the generated label is technically accurate but not how your team talks about the theme, rename it before classifying rather than living with an awkward label throughout your reporting.

Merging¶

Two categories that are really describing the same thing (common when a theme gets split into a slightly-too-fine distinction) can be merged into one. Merging keeps you from ending up with a report where three thin categories are really one story split three ways.

Splitting¶

The reverse problem also happens - a category that's technically correct but too broad to be useful, bundling several distinct themes under one label. Splitting a broad category into more specific ones before classifying gives you a report that reflects the actual variety in what people wrote, instead of one oversized bucket.

Adding a Category Manually¶

If you already know there's a theme you want tracked - even one the generated set missed, or one you know from prior research is worth watching for - you can add it directly rather than relying entirely on what the automated pass surfaced.

Deleting¶

A category that doesn't hold up on review (too narrow, redundant, or simply not useful for what you're trying to learn) can be removed before it's ever applied to a response.

When to Re-Run Classification¶

Any edit to your category set - a rename, merge, split, addition, or deletion - means the existing classification no longer reflects your current categories. Re-run classification after you're done editing, rather than after every individual tweak, so you're not spending credits on repeated partial runs.

Re-running classification reclassifies every response in the question from scratch, including any you've already manually corrected - a fresh run doesn't preserve prior manual corrections. If you've already done manual review on individual responses (see Reviewing and Correcting Individual AI Classifications), do that step after your categories are finalized and you're done running classification, not before - otherwise a later run resets it.

FAQ¶

Can I edit categories after responses have already been classified?
Yes - editing categories at any point is fine. Just remember that classification needs to be re-run afterward for the results to reflect your changes.

Is there a limit to how many categories I can have?
There's no hard cap, though a very large category set tends to become harder for both you and your readers to interpret than a tighter, more deliberate one - if you find yourself with dozens of categories, merging related ones is usually worth doing.

Can multiple people edit the same category set?
Category edits apply to the survey's shared Text Analytics results, so any changes are visible to anyone else with access to the survey.

Does editing categories cost credits?
No - editing the category set itself (renaming, merging, splitting, adding, deleting) is free. Credits apply to generating the initial category set and to running classification.

Once your categories are set, head back to How Text Analytics Works for the classification step, or straight to building a report from the results in Building Your First Report.