Review analysis
Open Analysis to explore imported Reviews by App, period, Store, version, severity, intent, and topic. Select a chart segment to inspect the Reviews behind it; open a Review to return to its Reply Inbox workflow.
Understand the results
Section titled “Understand the results”Jev classifies Reviews into shared intents, reported severity, and topics from the App’s catalogue. A Review can have several intents/topics or none. Luna proposes missing English topic labels and definitions for human approval; neither model publishes replies.
- Severity describes reported impact, not star rating or a verified incident. Critical means reported data loss or financial harm; a refund request alone is not critical.
- Unknown severity and unclassified Reviews are distinct from no problem.
- Pending topics appear provisionally with a visible label; rejected topics are excluded from active analysis.
- Counts represent unique Reviews in each slice. Topic counts overlap and cannot be added into a Review total.
The period uses the Store’s Review date. Version filtering requires an App and matches the label across both stores unless a Store is selected. Missing versions stay visible as an explicit state.
Correct a classification
Section titled “Correct a classification”All Organization members can correct a Review’s intents, severity, and topics, or return it to automatic classification. Corrections survive reclassification. When the Review changes, the correction is flagged for rechecking.
Manage the topic catalogue
Section titled “Manage the topic catalogue”Owners and admins can approve, reject, rename, reopen, merge, and create topics for an App. Both of its Store Connections share the catalogue.
- Renaming preserves the topic identifier and assignments.
- Merging requires a surviving topic; its status is retained and source labels become aliases.
- Rejecting requires confirmation and removes assignments and manual topic corrections. Reopening does not restore them.
- Rejected topics block equivalent proposals until reopened.
Availability
Section titled “Availability”Analysis covers all imported history in resumable batches; it cannot retrieve history absent from the Store API. Changed Reviews wait for fresh automatic analysis. Catalogue edits do not reclassify all previous Reviews.
The deployment operator enables Jev with TYPESAFE_API_KEY and Luna with the generative provider settings. Missing configuration leaves stored results readable. Classification and discovery are separate from Reply Draft quotas; discovery runs in bounded batches without a published commercial quota. See self-hosting for configuration.