Go or no-go: deciding a bad release in minutes

Release decisions go wrong slowly, then suddenly. The slow part is months without agreeing who decides. The sudden part is a red dashboard with six people waiting for someone else to speak. A go/no-go frame agreed in calm hours turns the sudden part into minutes.

Decide on two axes

Blast radius: growing or flat. Count affected users, endpoints or revenue paths, and note the trend. Growing beats big-but-flat every time.

Fix confidence: certain or uncertain. Certain means the fix is known, tested and minutes away. Everything else is uncertain, including fixes that look easy.

Growing plus uncertain means no-go. Stop, hold, communicate. The matrix decides before egos enter the room.

Worked example: a fictional Friday rollout

The context below is fictional. Fictional team ParcelTrack (fictional) rolls out a new checkout fee to ten percent of traffic. Errors climb on the new path within twenty minutes; the fix idea is untested.

The decider calls no-go at minute 25: rollout frozen, new path drained, one message with the next check in an hour. Monday: the fee logic fixed with tests, rollout resumes. The postmortem praises the early call, not the fix speed.

Checklist: a decision that holds

1. One named decider before the window opens.
2. Blast radius and confidence stated out loud.
3. No-go communicated once with the next check time.
4. The call reviewed afterwards without blame.

Related reading

• Hands on: Go or No-Go on a Bad Release.
• The first-hour frame: Incident response.