Ship on Fridays

Every team that bans Friday deploys is telling you something true about their infrastructure. They are not saying Friday is dangerous. They are saying recovery is slow, and Friday is when slow recovery hurts most.

The rule is a symptom

Consider what the rule actually protects against. A bad release goes out, and the people who understand it are gone for two days. That is only frightening when undoing the release is hard — when it means finding the previous commit, rebuilding it, and hoping the build is reproducible.

If restoring the previous release takes ninety seconds and no build, the calendar stops mattering.

What to fix instead

• Keep every release. Storage is cheaper than a weekend.
• Make rollback a single command, available to everyone on the team.
• Never make recovery depend on a rebuild succeeding.

The Google SRE book (opens in a new tab) makes the same argument at much greater length: reduce the cost of failure rather than the frequency of change. Teams that do this stop scheduling deploys around their fears.

The honest version

Some releases genuinely should wait — a database migration, a change to billing, anything where “undo” is not a real operation. Those deserve care on any day of the week. But the blanket Friday rule is rarely about those. It is about not trusting the button.

Fix the button.