Skip to main content
Change failure rate (CFR) measures the percentage of production deployments that result in a failure. It is a direct indicator of release quality — how often does shipping code break something?

What Periscope tracks

Periscope calculates CFR from your deployment events. The dashboard shows:
  • Overall failure rate as a percentage
  • Total deployments and failed deployments count
  • Failure rate by service to identify which services are most failure-prone
  • Recent failed deployments with details (service, time, pipeline URL)

DORA benchmarks

LevelBenchmark
Elite0-5%
High5-10%
Medium10-15%
Low15%+

How it is calculated

change_failure_rate = failed_deployments / total_deployments * 100
Periscope counts deployments to production with status: "failure" against the total number of production deployments (any terminal status) in the selected time range. The per-service breakdown applies the same formula scoped to each service name in your deployment data.

Interpreting the data

  • Low CFR with high deployment frequency is the ideal state — you are shipping often and reliably.
  • High CFR with low deployment frequency is the most dangerous pattern — infrequent, large deployments that often break.
  • One service with high CFR while others are healthy suggests a problem specific to that service’s test coverage, complexity, or deployment process.
  • CFR trending upward may indicate degrading test quality, increasing system complexity, or a growing gap between development and production environments.

Reducing change failure rate

  • Ship smaller changes more frequently (smaller blast radius)
  • Improve pre-merge CI — catch failures before code reaches production
  • Add canary or staged rollout strategies
  • Implement feature flags to decouple deployment from release
  • Review failed deployments in Periscope to identify patterns

MCP tool

Query change failure rate from your AI coding assistant:
get_change_failure_rate(time_range: "30d")
Returns total deploys, failed deploys, failure rate percentage, per-service breakdown, and the 10 most recent failures.

Mean time to recovery

When failures happen, MTTR measures how fast you recover.