## Definition
**DORA metrics** are the four key performance indicators identified by the DevOps Research and Assessment programme (Nicole Forsgren, Jez Humble, Gene Kim — see [[Accelerate (Forsgren Humble Kim)]]) that empirically predict organisational performance.
## The Four
1. **Deployment Frequency** — how often the team releases to production.
2. **Lead Time for Changes** — time from commit to running in production.
3. **Mean Time to Restore (MTTR)** — how fast service is restored after an incident.
4. **Change Failure Rate** — % of changes causing a degraded service / rollback / hotfix.
## Velocity vs Stability
- **Velocity:** Deployment Frequency, Lead Time for Changes.
- **Stability:** MTTR, Change Failure Rate.
High-performing teams excel on **both axes simultaneously** — the metrics do not trade off when the system is well-designed.
## Why They Survive the Agentic Era
Old velocity metrics (story points, PR count, lines of code) are easily gamed by AI agents. DORA's four measure *delivered outcomes* — they hold meaning regardless of who wrote the code.
## The New Companion Metric
**Cost per merged PR** (tokens, dollars). New, agentic-era specific. Trend matters more than absolute number.
## Anti-metrics
Don't use:
- Lines of code (any direction).
- PR count.
- "AI-assisted code %" (almost everything is now).
- Self-reported "time saved" estimates.
## Practical Cadence
Weekly snapshot. Publish to a wiki / Slack canvas / repo `dashboard.md`. The act of publishing changes behaviour the first month; the trend tells you whether the system is improving the second and third months.
## Related
- [[Orchestrator Role]]
- [[Decommission Discipline]]
- [[Accelerate (Forsgren Humble Kim)]]