Cluster scattered feedback into themes

Turn fragments from comments, tickets, and interviews into a few themes you can act on

You are a user research analyst. Cluster the scattered feedback below into themes: 1) read it all first and cluster by what the user was trying to accomplish, not by keyword or sentiment; 2) for each theme give one summary sentence, the number of supporting items, two representative quotes verbatim, and one counterexample; 3) rank by reach times severity and state the basis for the ranking; 4) list the items that fit nowhere and any likely misreadings. Do not soften the wording or complete the users' logic for them.

[paste the comments, tickets, or interview notes]

How to use

Read the 'fits nowhere' list before you accept the themes. That is usually where the real anomaly is hiding.

Good for

Eighty comments and thirty tickets came in after a launch, the prioritization review is next week, and you need three to five themes that map directly onto changes.
The interview notes contradict each other, and you want to see what most people want and why a minority does not, rather than being led by the last person you spoke to.

Not for

There are only two or three pieces of feedback: clustering invents patterns that are not there, and reading and answering each one is more reliable.
You need to reply to each item or apologize: themes erase the case-level detail, and that is exactly what this situation requires.

Example

Input

Export keeps failing, I click and nothing happens. / Can you add dark mode? / Export spun for five minutes and then errored. / The table text is tiny on my phone.

What you get

It groups the two export failures as 'export gives no feedback and times out', keeps 'I click and nothing happens' as a verbatim quote, and separates dark mode and mobile table text into their own themes. The ranking states that it multiplies frequency by whether the task is blocked, not by how angry the comment sounds.

Common mistakes

Clustering by keyword: every mention of 'export' lands in one bucket, so 'export failed' and 'export format is limited' merge into a theme with two different fixes.
Forcing every item into a theme: the few that fit nowhere are often the ones the product most easily ignores and most needs to see on their own.

Try tweaking it like this

Replace 'reach times severity' with 'effort divided by reach' when resources are tight and you need the cheap, high-leverage items first.
Add 'for each theme, one line on what happens if this judgment is wrong' to keep a prioritization meeting from treating a hypothesis as a finding.