Status pages have a credibility problem: all systems operational, says the banner, while your dashboard times out. Beacon rebuilt Prism's incident tooling around a blunt principle — the status page reads from the same probes the on-call engineer does, with no editorial layer in between.
The honesty architecture
Beacon's status is computed, not declared. Synthetic probes hit real user flows — login, checkout, API reads and writes — from five regions, and the public page renders whatever they measure. An incident can be annotated with context, but the underlying signal can't be edited. Uptime numbers link to the raw probe data, and the 90-day history can't be quietly truncated.
Designing the worst moment
People visit a status page at the worst moment of their day with your product. The design brief was a calm emergency room: current state answers first, in plain language, above the fold, on a page that weighs 40KB and renders without JavaScript — because when your CDN is the incident, the status page still has to load.
- Time-to-first-status under 200ms globally, fully static delivery
- Incident timeline writes require two keystrokes from the on-call terminal
- Subscriber notifications fan out within 30 seconds of state change
The uncomfortable launch
The first month was rough on purpose: computed status meant Prism's real uptime — 99.7%, not the 99.99% the old page claimed — was suddenly public. Sales was nervous. Then a major incident hit, the page told the truth in real time, and three enterprise customers cited the transparency in renewal calls. Honesty turned out to be a feature you can sell.