Reliability

Deployment reliability over a 7d / 30d / 90d window, for the cluster and per workload.

The four numbers

Stat	What it means
Deploys	Total deployments in the window, and how many workloads they touched.
Change-Failure Rate	Share of deploys that failed or were rolled back.
Fragile workloads	Workloads whose own CFR is ≥ 30%.
Median MTTR	Median time from a failed deploy to the next successful one.

Change-failure rate and MTTR are two of the four DORA metrics, computed from deploy history the agent observed rather than from a CI system’s own reporting — which means they count the rollbacks CI never heard about.

Per-workload timeline

Below the stats, each workload gets a deploy timeline: one mark per deploy, coloured by outcome, with a release verdict chip. INSUFFICIENT_DATA is a real verdict here — a workload with two deploys in ninety days has no meaningful failure rate, and the tab says so rather than printing a confident 0%.

Gate accuracy

The last card scores Runtimez against itself: when the risk gate predicted a release would fail, was it right? Precision and recall over recorded predictions. Until predictions have been recorded it shows —, not a flattering default.

From the CLI

rtz changes --since 168     # the deploy history behind these numbers