dora-metrics

DORA Metrics Knowledge

This skill provides knowledge about DORA (DevOps Research and Assessment) metrics, performance bands, and how git patterns correlate with delivery performance.

---

The Four Key Metrics

DORA research (Forsgren, Humble, Kim — Accelerate, 2018) identified four metrics that reliably predict software delivery performance and organizational outcomes:

1. Deployment Frequency

How often code is deployed to production.

Performance Band	Frequency
Elite	On-demand, multiple times per day
High	Between once per day and once per week
Medium	Between once per week and once per month
Low	Less than once per month

Git proxy: Commit-to-main frequency, tag/release cadence.

2. Lead Time for Changes

Time from code commit to code running in production.

Performance Band	Lead Time
Elite	Less than one hour
High	Between one day and one week
Medium	Between one week and one month
Low	More than one month

Git proxy: Branch lifespan, PR open duration, merge-to-deploy interval.

3. Mean Time to Restore (MTTR)

Time to recover from a production failure.

Performance Band	MTTR
Elite	Less than one hour
High	Less than one day
Medium	Between one day and one week
Low	More than one week

Git proxy: Time between revert commit and fix commit, hotfix frequency.

4. Change Failure Rate

Percentage of deployments that cause a failure in production.

Performance Band	Failure Rate
Elite	0–15%
High	0–15%
Medium	16–30%
Low	>30%

Git proxy: Revert rate, hotfix branch frequency, rapid-fix-after-deploy pattern.

---

Key Research Findings

The core insight from a decade of DORA research:

Teams with short-lived branches (under 24 hours), daily commits to trunk, and small batch sizes consistently outperform those practicing feature branching — by margins of 182x faster deployment frequency and 127x faster lead times.

Performance Multipliers

Practice	Elite vs Low Performance Gap
Deployment frequency	182x more frequent
Lead time for changes	127x faster
Mean time to restore	2,604x faster
Change failure rate	7x lower

These metrics are correlated, not independent:

• Teams that deploy more frequently have lower failure rates (not higher).
• Faster lead times correlate with faster recovery.
• Throughput and stability reinforce each other.

---

Git Patterns and DORA Correlation

Patterns That Correlate with Elite Performance

Git Pattern	DORA Impact	Why
Trunk-based development	Elite deployment frequency	Eliminates merge overhead, enables continuous flow
Branches < 24 hours	Elite lead time	Short feedback loops, minimal integration debt
Small commits (< 200 lines)	Lower change failure rate	Easier to review, test, and bisect
Daily commits to main	Elite deployment frequency	True continuous integration
Automated merge (CI green → deploy)	Elite lead time	Removes manual bottleneck
Feature flags over branches	Elite all metrics	Decouples deployment from release

Patterns That Correlate with Low Performance

Git Pattern	DORA Impact	Why
Long-lived feature branches (> 7 days)	Low lead time	Integration debt accumulates
Branch-per-environment	Low deployment frequency	Violates "build once, deploy everywhere"
Cherry-pick promotion	Low change failure rate	Different artifacts in each environment
GitFlow with long-lived branches	Medium at best	Ceremony overhead discourages small batches
Mandatory squash merging	Impairs MTTR	Destroys bisect capability
Infrequent commits (weekly)	Low deployment frequency	Batching is the opposite of flow
Code freeze periods	Low deployment frequency	Elite teams never freeze
High revert rates (> 15%)	Low change failure rate	Indicates quality pipeline failures

---

Trunk-Based Development Benefits

Bryan Finster's Minimum CD principles and Martin Fowler's CI definition both require:

1. Integration at least daily — branches should be short-lived (< 24 hours ideal)
2. Commit to trunk — not to long-lived feature branches
3. Always green trunk — main branch should always be deployable
4. Feature flags — use flags to hide incomplete features, not branches
5. Small batches — frequent, small changes over large, infrequent ones

Why Trunk-Based Works

• 50%+ merge conflict probability when branches live > 3 days
• Integration failure guaranteed when branches live > 30 days
• Merge conflicts scale quadratically with branch count and age
• Code review quality degrades with PR size (research: > 400 lines = diminishing returns)

---

Using DORA in Audits

When performing a git health audit:

1. Measure the proxies — branch age, commit frequency, revert rate, release cadence
2. Map to DORA bands — Elite / High / Medium / Low
3. Identify the bottleneck — which metric is the weakest?
4. Prioritize by impact — improvements to the weakest metric yield the largest gains
5. Track over time — DORA metrics are trend indicators, not point-in-time scores

References

• Forsgren, N., Humble, J., Kim, G. (2018). Accelerate: The Science of Lean Software and DevOps
• DORA State of DevOps Reports (2014–2024)
• Finster, B. et al. "Minimum CD" — minimumcd.org
• Fowler, M. "Continuous Integration" — martinfowler.com
• Humble, J., Farley, D. (2010). Continuous Delivery