DORA Four Metrics

The DORA Four Metrics are four empirically validated measures of software delivery performance, developed by the DevOps Research and Assessment (DORA) program — a multi-year research effort led by Nicole Forsgren, Jez Humble, and Gene Kim. First published in Accelerate (2018), they are the most widely cited outcome-based framework for assessing engineering team health.

The Four Metrics

MetricWhat It MeasuresElite Benchmark
Deployment FrequencyHow often code reaches productionOn demand (multiple times/day)
Lead Time for ChangesTime from commit to production< 1 hour
Change Failure Rate% of deployments causing incidents0–15%
Time to Restore ServiceRecovery time after a production failure< 1 hour

Performance Tiers

  • Elite: Deploy on demand; lead time <1hr; change fail rate 0–15%; MTTR <1hr
  • High: Deploy 1/day–1/week; lead time 1 day–1 week; change fail rate <15%; MTTR <1 day
  • Medium: Deploy 1/week–1/month; lead time 1 week–1 month
  • Low: Deploy 1/month–6 months; lead time 1 month–6 months

Why These Metrics Resist Gaming

Each metric has a natural counterbalancing partner, making it difficult to optimise one in isolation:

  • Deployment FrequencyChange Failure Rate: Deploying more often could raise failures — elite performers achieve both simultaneously, proving the metrics are genuinely linked to capability, not process tricks.
  • Lead TimeTime to Restore: Rushing changes through (shortening lead time) tends to increase incidents; if MTTR is poor, teams self-limit release speed. The combination reveals true pipeline health.

Because each metric naturally constrains its counterpart, gaming requires improving the underlying system — which is the point.

Larson’s Application

Will Larson treats the DORA metrics as the anchor for engineering measurement in two key contexts:

  • Four-States-of-a-Team: A team “Falling Behind” exhibits worsening deployment frequency and lead times; a team being “Disrupted” by technical debt shows rising change failure rates.
  • Directional Metrics: Larson recommends DORA as the top-level health indicators for any engineering organisation, used alongside team-level directional metrics to track trajectory over time.

Recent Developments

The 2021–2023 DORA research added Reliability (operational health and service availability) as a fifth emerging dimension, reflecting the growing importance of platform and SRE thinking in high-performing teams.

Sources

  • Forsgren, Nicole, Jez Humble, and Gene Kim (2018). Accelerate: The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations. IT Revolution Press. ISBN: 978-1-942788-33-1.

    • Original publication of the DORA Four Metrics and performance tier benchmarks; based on four years of survey research with 23,000+ respondents.
  • Larson, Will (2019). An Elegant Puzzle: Systems of Engineering Management. Stripe Press. ISBN: 978-1-7322651-8-9. Chapter 2.4.

    • Application of DORA metrics to engineering team states and directional measurement.
  • DORA / Google Cloud (2023). 2023 State of DevOps Report. Google LLC. Retrieved from https://cloud.google.com/devops/state-of-devops

    • Annual update to the DORA research, confirming core four metrics and introducing reliability as an emerging fifth dimension.
  • Forsgren, Nicole, Dustin Smith, Jez Humble, and Joanne Molesky (2019). “The SPACE of Developer Productivity.” Communications of the ACM. Discussion of how DORA metrics complement broader developer productivity frameworks.

    • Peer-reviewed academic validation of the metrics’ correlation with organisational outcomes (profitability, market share, goal achievement).
  • Humble, Jez and David Farley (2010). Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation. Addison-Wesley. ISBN: 978-0-321-60191-9.

    • Theoretical foundations of deployment frequency and lead time as pipeline health signals, predating the formal DORA research.

Note

This content was drafted with assistance from AI tools for research, organization, and initial content generation. All final content has been reviewed, fact-checked, and edited by the author to ensure accuracy and alignment with the author’s intentions and perspective.