Does our radar see what companies declare?
We compare the radar's outage detections with the incidents that companies declare on their own status pages. A product that measures others should measure itself first.
Loading the latest results…
Recall · critical + major—no cases yet
Precision · DOWN episodes—No covered service had an outage episode.
Median lead vs declaration—Needs matched incidents.
Definitions
Rule val-v1. Ground truth: incidents from — machine-readable status pages, re-read every ten minutes.
| Term | Definition |
|---|---|
| DOWN episode | Two or more consecutive failed five-minute sweeps of a service. Sweeps in which more than half of all probes fail are excluded (our own network). |
| Detected | An official incident counts as detected if any monitored service of that company had a DOWN episode between 30 minutes before the incident started and 30 minutes after it was resolved. |
| Precision | Share of DOWN episodes of covered services that overlap an official incident of the same company. |
| Covered service | Reachable in at least half of all sweeps (bot walls and geo blocks are not outages) and run by a company with a status page. |
Recall by official impact
Window — to —.
| Official impact | Detected / declared | Recall | 95% CI |
|---|---|---|---|
| critical | — | — | — |
| major | — | — | — |
| minor | — | — | — |
| none | — | — | — |
Wilson score intervals. A recall of 0 with a small sample is still informative: the upper bound says how good the radar could plausibly be.
Severe incidents the radar missed
Listed, not hidden. Each links to the company's own incident page.
| Declared | Company | Impact | Incident |
|---|---|---|---|
| No severe incidents in the window. | |||