Clusters & triage

Clusters are themes — groups of feedback that mean the same thing. Sentriment embeds every item and groups by meaning, not keywords, so “the app is slow” and “takes forever to load” land together while “slow to respond to my support email” does not. Themes emerge on their own, grow as feedback arrives, and reopen automatically if an issue you marked resolved comes back.

Each theme carries a label, a representative summary, a count, a sentiment mix, and its revenue at stake. You never create or maintain them by hand.

The Clusters page is built for deciding what to do, not just browsing. It has several views:

• Triage — the worklist: what's new, spiking, or biggest, with pin / mute / resolve actions.
• Areas — themes grouped into higher-level areas so you can brief by product surface.
• All — the full list, sortable by volume (with dates), recency, or revenue at risk.

Expand any theme to read its feedback, see which segments over-index in it, and file it to GitHub or Linear with the quotes attached. When two themes are really one, Sentriment suggests a merge with a similarity score.

A theme that suddenly surges past its own baseline is a spike — your early-warning system for a breaking issue. A theme you marked resolved that starts collecting fresh reports is a regression— the fix didn't stick, or a release broke it again. Both can notify you the moment they happen; see Alerts.

When your users carry an mrr trait, each theme shows the total MRR of the users complaining in it — so “dark mode” and “billing is broken” stop looking equally urgent. This is the total for the theme, not a windowed slice.