Deploy Frequency Target

Daily deploys. The target.

DORA frequency targets

DORA's four tiers anchor the conversation: elite ships multiple times per day, high ships weekly to daily, medium weekly to monthly, low below monthly. The headline result is that elite performers also have the lowest change failure rate; frequency and reliability are correlated, not traded off.

• Elite tier. Multiple deploys per day, change failure rate under 15%, restore-from-failure measured in under an hour.
• High tier. Weekly to daily deploys with comparable failure metrics; strong but with headroom for automation gains.
• Medium and low tiers. Weekly to monthly or slower; this is catch-up territory where toil and approval theatre dominate the schedule.
• Frequency correlates with reliability. Elite teams also lead on change failure rate and MTTR; small batches, fast feedback, and reversibility compound.

Setting your target

Set the target from your current baseline, not from the elite tier on a slide. A 2x improvement in six months is the stretch goal that surfaces the real blockers without inviting cargo-cult automation.

• Pull a 90-day baseline. Count actual deploys per service per day from your CI history; this is the only honest starting line.
• 2x in six months. A doubling target forces conversations about CI duration, approval flow, and coupling that vague goals do not.
• Do not skip tiers. Low to elite in one cycle requires capability the team does not yet have; aim for the next tier.
• Document the named cadence and the deadline. "We said we would improve" rots into nothing without a number and a date.

Measure honestly

Measure per service per day, not aggregate. A 100-service monorepo deploying once per service per week is not elite even though the topline number looks busy.

• Per service per day. Aggregate counts hide the services that have not deployed in months.
• Exclude config-only deploys consistently. Whichever rule you pick, apply it the same way every quarter.
• Team-owned dashboard. Visibility shifts behaviour; the team that watches its own number improves it without needing edicts.
• Quarterly methodology audit. Data sources drift; rebuild the dashboard from raw events on a fixed cadence to catch silent regressions.

Common blockers

Three blockers absorb most of the gap to elite: approval theatre, slow CI, and coupled deploys. Each halves the achievable rate; tackle them in that order if you have to pick one.

• Approval theatre. Three or four manual sign-offs per deploy halve the rate; cut to one risk-based approval and automate the rest.
• Slow CI. Pipelines over 30 minutes turn deploys into events; sub-10-minute CI is the precondition for everything else.
• Coupled deploys. A monolith that gates every release is the dominant blocker once CI is fast; decouple the painful surface first.
• Documented per-blocker impact. Measure the rate change after each fix so the next quarter's prioritisation has data.

How to reach the target

Get there with weekly tracking, a single blocker focus per quarter, and the discipline to refuse to declare victory on frequency alone. Frequency without low CFR and fast MTTR is just noise.

• Track weekly. Publish the trend, not just the absolute number; direction matters more than any single week.
• One blocker per quarter. Name it, knock it down, measure the lift. Three concurrent initiatives finish none.
• Frequency with health. Pair every frequency report with change failure rate and MTTR; refuse to celebrate one without the others.
• Published scoreboard. Visible CFR, MTTR, and frequency together force honesty and direct the next quarter's focus.