Download the incident dataset
incidents.csv holds all 40,026 incidents on this site, one row each, rebuilt every three hours (8.8 MB). Each quarterly report has its own file covering that quarter.
It is licensed CC BY-NC 4.0: free to reuse for research, journalism and other non-commercial work, with credit to sevanalysis. Commercial use needs a separate license: write to hello@sevanalysis.com. The methodology explains how the figures are made and what they cannot tell you.
Columns
| Column | Meaning |
|---|---|
| vendor | Short vendor id, as in the vendor's page address on this site |
| vendor_name | Vendor name |
| tier | infra, app or ai |
| id | The incident's id on the vendor's status page |
| title | Title as the vendor posted it |
| status | Latest status the vendor posted, such as investigating or resolved |
| impact | The vendor's rating: none, minor, major or critical, or unknown where the status page has no rating |
| started_at, resolved_at | UTC, ISO 8601. resolved_at is empty while the incident is open |
| duration_minutes | Start to resolution. Empty while open |
| mitigation_minutes | Start to the first "monitoring" update. Empty for vendors that never post that stage |
| repeat | 1 when the vendor posted the same title within the previous hour; counted once in the figures |
| components | Affected components, separated by ; |
| mentions_upstream | Providers the incident names, separated by ; |
| failure_mode, root_cause | Labels from the taxonomy. Empty for open incidents and most older ones: labelling covers resolved incidents from the last four months |
| url | The incident on the vendor's status page |