Feedback
Rate the accuracy of an analysis to drive your team's quality metrics and improve future diagnoses
Every complete analysis can be rated. Feedback shows up in your team's quality metrics over time and helps SigSentry surface more relevant similar incidents.
How to rate
Analysis detail page → Rate this analysis at the bottom. Pick one:
| Rating | Meaning |
|---|---|
correct | Root cause and suggested actions were accurate |
partially_correct | Got the general area right, missed specifics |
incorrect | Wrong service, wrong cause, or actions weren't useful |
Optionally add a comment explaining what was right or wrong. The comment is stored alongside the analysis and visible to anyone with permission to see the analysis.
When to rate
The most useful time to rate is after you've resolved the incident — at that point you know whether the diagnosis matched reality. Earlier ratings are still helpful but more speculative.
A rating doesn't change the analysis — the diagnosis stays as-is.
Why rating matters
Drives team quality metrics
Your dashboard's Usage page shows accuracy by month, by category, and by trigger source.
Improves similar-incident matching
Rating analyses helps SigSentry surface more relevant past incidents on future diagnoses.
Surfaces problem categories
If your team consistently rates a particular error category as incorrect, it's a signal to add detail to your project's AI context.
Where ratings show up
| Surface | Shows |
|---|---|
| Analysis detail | The rating + comment, plus who rated |
| Activity log | Filter analyses by rating |
| Usage dashboard | Aggregate accuracy over time, by month and category |
| Similar incidents | Rated entries help shape future match suggestions |
Changing a rating
The user who rated can re-rate from the same control. Other users with analysis-write permission can also override. The audit log records who changed the rating and when.
Bulk rating
For teams catching up on backlogged analyses:
- The analysis list view has an Unrated filter
- Multi-select + "Mark as correct" / "Mark as incorrect" applies a rating to all selected at once
- Useful right after onboarding, when your team accumulates a few weeks of analyses without rating any
Privacy
Ratings and comments are visible to your team only. They're never shared with other tenants and never used to train models that affect other customers.