Similar incidents
Surface past analyses that look like the current one — useful for "have we seen this before?" and tracking recurring issues
Every complete analysis includes a Similar incidents card listing prior analyses that resemble the current one. Useful for spotting recurring issues, finding the previous fix, or noticing patterns the team has already triaged.
Similar incidents is a Pro+ feature. Starter accounts see the section but with no entries.
Where to find it
In the dashboard, the analysis detail page has a Similar incidents card below the suggested actions. It shows up to 5 matches by default, expandable to 20.
Each entry includes:
| Field | Notes |
|---|---|
| Title | The matched analysis's summary headline |
| Date | When that analysis ran |
| Similarity | A 0–1 score for how close the match is |
| Severity | The matched analysis's severity |
| Resolution status | Whether feedback was logged on it |
Click any entry to jump to that analysis.
How matching works
Each entry has a similarity score between 0 and 1 — higher means more alike. Recent incidents and ones your team has rated tend to surface first. Raw log content is never compared.
Use cases
Find the previous fix
An incident that resembles a past one — open the past analysis to see what action was suggested and whether feedback indicated it worked.
Detect recurring issues
If five analyses of the same service all match each other, the underlying issue keeps coming back. Time to fix it for real.
Confirm a hypothesis
If you suspect this is the same thing that happened during last Tuesday's deploy, look at the similar-incidents list to confirm.
Tuning the matches
If the matches feel off-base:
| Symptom | Try |
|---|---|
| Too few matches | The current analysis is genuinely novel, or your account hasn't accumulated enough history yet |
| Many matches but they're all unrelated | Description is too generic — "errors" matches everything. Use a more specific description and re-run |
| Matches all from one team / project | Expected if you scope to a single project — the AI's pool is what your tenant has run |
Privacy
Similar-incident matching is scoped to your tenant. Analyses across tenants are never compared. A match never reveals data from another tenant.
What gets compared
Only analysis metadata (root cause, services, error type, severity). The raw log evidence is not used in matching — privacy and performance both benefit from this.