Deploy Windows vs Continuous: When Each Is Right

Continuous deploy is the engineering ideal; deploy windows are the realistic compromise for many businesses.

The continuous-deploy case

Continuous: code merges → CI passes → production. No human gate.

Optimizes for speed; assumes mature observability and rollback.

The deploy-windows case

• Deploy windows: deploys only during specific hours; manual approval often required.
• Optimizes for low-risk windows; assumes deploy carries non-trivial risk.

Four signals

1. Customer base sensitivity. Banks; healthcare; high-stakes.

2. Team maturity. Without rollback discipline, continuous deploy is unwise.

3. Observability depth. Continuous deploy needs strong observability.

4. Regulatory. Some regimes prohibit silent prod changes.

The hybrid posture

Many teams: continuous deploy to staging; window-gated deploys to prod.

Or: continuous deploy to prod; manual gate for high-risk changes only.

Antipatterns

• Continuous deploy without observability. Outages without diagnosis.
• Deploy windows that span 4 days. Defeats the purpose.
• Manual gate as the only safeguard. Replace with metrics + automation.

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.