Increasing Transparency on GitHub's Status Page

GitHub is introducing three changes to increase transparency on its service status page. It is making incident classification more granular, publishing 90-day uptime for each service, and separating Copilot's model provider outages into a distinct component.

First, a new severity tier called Degraded Performance is being added. This refers to a state where a service is functioning but experiencing delays, reduced functionality, or intermittent errors on some requests, operating as a 3-tier system alongside the existing Partial Outage and Major Outage.

This change reduces the limitation of the previous classification, where even minor issues appeared as partial outages. GitHub explains that situations where a service is only partially inconvenienced, as before, can now be displayed more accurately in line with actual user impact.

The status page will directly display the 90-day uptime percentage for each service. The calculation method reflects the severity and duration of incidents, following industry-standard status page calculation methods.

The key weightings are as follows.

• Major Outage: 100% of total downtime is counted as downtime
• Partial Outage: only 30% is reflected as downtime
• Degraded Performance: treated as 0%, considered as the service still functioning

For example, if a 1-hour Partial Outage occurs over 90 days, only 18 minutes of effective downtime is reflected in the uptime calculation, rather than the full hour. Degraded Performance incidents, on the other hand, do not affect the uptime figure.

Copilot-related notices are also becoming more refined. Previously, when a model provider outage occurred, it was recorded as an overall Copilot incident, but going forward it will be separated into a Copilot AI Model Providers component to more clearly convey which model is affected.

This change is particularly meaningful for features that support multiple models, such as Copilot Chat and GitHub Copilot cloud agent (formerly coding agent). Even if one model experiences an outage, users can select a different model or utilize auto model selection, so the purpose is to more accurately reflect the actual user experience.

GitHub emphasizes that clear communication during an outage is most important. These changes are measures to increase transparency, accuracy, and speed so that operations teams can make better decisions.