Methodology · v3.1

USTHistorique d'uptime

Observed 90-day uptime for {legal_name}'s public-facing status page. EXTRACTION RULES: Locate the vendor's status page (status.<vendor>.com, <vendor>.statuspage.io, third-party status aggregators). Cite the page + third-party status-history aggregator if available + an independent network-outage observability source for major outage corroboration. Return exact_numerical{value: <uptime_pct>, unit: "percent", source_urls: [status page + corroborator]}. The value MUST be an uptime PERCENTAGE in [0,100] read from the status page (e.g. 99.95). DO NOT return downtime minutes, a basis-point figure (9999), or 0 as a parse-failure sentinel — a value outside [0,100] (or a 0 that is not an affirmatively-evidenced full outage) is garbage and MUST be returned as type="absent" reason="no_reliable_uptime_value", NOT mapped into the bottom band. Server applies ratio_band (99.99+→95 four-nines, 99.9-99.99→85 three-nines, 99-99.9→70, 95-99→45, <95→20). If the vendor publishes NO public status page, return type="absent" reason="no_public_status_page" — DO NOT fabricate uptime from absence-of-news.