Deploy Frequency as the Master Metric for Engineering Health
The DORA research is consistent: deploy frequency is the metric that correlates most with healthy engineering. Move it; everything follows.
Why frequency over volume
Frequent deploys are smaller. Smaller deploys are easier to reason about. Easier reasoning leads to faster recovery and lower failure rate.
The metric is causal, not correlated. Move it intentionally and the rest of the system gets healthier.
Four practices that move it
- 1. Trunk-based development.
- 2. Feature flags.
- 3. Automated tests with high coverage.
- 4. Progressive delivery (canary).
When the metric lies
Counts ‘deploys per day’ without weighting by impact: a config-only deploy and a feature deploy count the same.
For DORA fidelity, count deploys that change application code; exclude config-only.
Healthy targets
Elite: multiple per day. High: between once-per-day and once-per-week. Medium: weekly to monthly. Low: monthly or less.
Most teams self-report higher than reality. Audit honestly.
Antipatterns
- Optimising for the metric without changing process. You can game it; you cannot fake the benefits.
- Measuring without context. A 50-engineer team with 3 deploys/day looks bad next to a 5-engineer team with 5/day.
- Annual reporting only. Quarterly is the minimum cadence.
What to do this week
Three moves. (1) Apply this to one pipeline first. (2) Measure deploy frequency / MTTR before/after. (3) Document the outcome so the next team starts from data.