Checks, locations & failure confirmation
How this is scored
What can be monitored and how reliably a failure is detected: HTTP(S) with status code and keyword checks, TCP and port, ping, DNS records, SSL certificate and domain expiry, cron jobs and heartbeats, multi-step API checks with assertions, browser or transaction checks of a login or checkout, the shortest check interval, the number and regions of probe locations, the re-check from a second location before an alert, timeouts and response time thresholds, and private locations inside the customer's own network.
0 — Runs no checks of its own; a status page or alerting tool that only receives signals from other monitors.
3 — HTTP checks of a URL at a fixed interval with an up or down result; keyword, SSL, cron and the probe locations are basic or not described.
5 — HTTP(S), keyword, TCP or port, ping and SSL expiry checks, cron or heartbeat monitoring, an interval of one minute or less, several named probe locations, a failure confirmed from a second location before alerting, and response times recorded per check.
8 — Multi-step API checks with assertions on headers and body, browser or transaction checks of a user flow, DNS and domain expiry checks, sub-minute intervals, probe locations listed by city with the confirmation logic documented, private locations or agents for internal services, and maintenance windows that pause checks.
10 — Coverage and accuracy a reliability team runs on: every check type above with code-based checks (for example Playwright or scripted tests) in version control, location-specific thresholds, false-positive handling and probe health published, and the detection method documented so a result can be reproduced.
The SaaS Platform Engineer
The check catalogue reads like my weekly backlog: HTTP with keyword and status-code conditions, TCP and UDP ports, ping, DNS, mail servers, SSL with chain, protocol and hostname validation, domain expiry, heartbeats driven by five-field cron expressions, multi-step API request chains and recorded browser transactions replayed from a chosen probe, on intervals down to 30 seconds. Failure confirmation is exactly what I want before anyone is paged — one to three backup probes must agree before an incident opens, recovery needs a single passing check, and the checkpoint directory lists every location with city and IPv4 and IPv6 addresses. What holds it below the top band is that we found no public information on scripted or code-based checks kept in version control, or on alert limits tuned per location. 2 2 4 12 13 21