Technology: DORA Metrics

Metrics make delivery speed and stability visible

DORA (DevOps Research and Assessment) has identified, through multi-year research, metrics that reliably distinguish high-performing from low-performing software teams. They measure not busyness but delivery efficiency and stability.

DORA metrics are a mirror, not a goal. Teams that treat them as KPIs with fixed targets optimise the metrics rather than improving the underlying system. The model has evolved from the original four keys to the current five-metric model.

The Five DORA Metrics

Throughput:

Change Lead Time: Time from committing to version control to deployment in production.
Deployment Frequency: How often is code deployed to production?
Failed Deployment Recovery Time: Time to recover from a deployment that fails and needs immediate intervention. Replaces the earlier Mean Time to Recovery (MTTR).

Stability:

Change Fail Rate: What share of deployments requires immediate intervention, such as a rollback or hotfix?
Deployment Rework Rate: What share of deployments is unplanned and happens as a result of a production incident?

Historical note: Early DORA reports grouped teams into clusters (Elite to Low) with fixed thresholds, for example deployment frequency multiple times per day versus monthly, or change failure rate 0-5% versus above 30%. These cluster boundaries come from the original four-key model and should be read as historical orientation, not as currently valid benchmarks. Today's positioning uses the DORA Quick Check and the respective dated report.

The Cluster Finding

High-performing teams share two characteristics: high deployment frequency and a low change fail rate. The widespread assumption that speed comes at the cost of stability is contradicted by the data.

Application

Measure a baseline: Establish current values without setting targets.
Identify the bottleneck: Which metric deviates most from the desired level?
Improve systemically: Improve the process, not the metric. Example: a long lead time points to manual pipeline steps, long review wait times, or missing automation.

Focus: Make System Bottlenecks Visible

The metrics show symptoms. The root causes lie in processes, tooling, and team structures.

FAQ

How do we capture DORA metrics without expensive tooling?

Deployment Frequency: count Git tags or deployment logs. Change Lead Time: commit timestamp versus deployment timestamp. Failed Deployment Recovery Time: incident ticketing system. Change Fail Rate: rollbacks or incident tags in deployments.

Should we include DORA metrics in OKRs?

With caution. As orientation values, yes. As a performance measure for teams it is risky, because it invites gaming.

References

DORA Software Delivery Performance Metrics. Current five-metric model and Quick Check. (2024). dora.dev/guides/dora-metrics/
Forsgren, Humble, Kim Accelerate: The Science of Lean Software and DevOps. Research foundation for the DORA model. (2018). www.oreilly.com/library/view/accelerate/9781457191435/
DORA State of DevOps Report. Annual industry benchmark. (2024). dora.dev
DORA Quick Check. Self-assessment tool for the current five-metric model. (2024). dora.dev/quickcheck/

Ask AI

These links open external AI services, the conversation and its content are sent to their providers.