I Deploy on Fridays (And Maybe You Should Too)
summary
Argument says deploying Friday is fine if a repeatable pipeline, tests and rollout controls exist: continuous deployment with a 15 minute pipeline, automated checks, rolling updates and feature toggles. Concrete patterns include a strangler proxy and branching-by-abstraction to carve legacy code, a staged feature-flag rollout (10% → 50% → 100%) that improved search metrics, and trunk‑based development. Research numbers back it: 46× deployment frequency, 100× faster recovery and 1/5 change‑failure rate for high performers.
AI-written summary · about
chapters (31)
- 0:00Start
- 1:41DEPLOYING?
- 2:50TRADITION
- 4:16CONTROL
- 5:17CONFIDENCE & TRUST
- 6:02TIME TO REPAIR / RECOVER
- 8:47WHAT ABOUT BUGS?
- 11:42BIG STEPS
- 12:02SMALL STEPS
- 12:16MAKE IT EFFORTLESS
- 12:52REDUCE RISK
- 20:57DEALING WITH "LEGACY"
- 21:30STRANGLER PATTERN
- 26:29CONFLICTS
- 27:09CONTINUOUS INTEGRATION
- 28:11DELAYING INTEGRATION
- 29:05FEATURE TOGGLES
- 31:10FEATURE TOGGLE DEBT
- 31:33BRANCHING STRATEGIES
- 32:05TRUNK BASED DEVELOPMENT
- 32:59SHORT LIVED FEATURE BRANCHES?
- 34:06PULL REQUESTS?
- 34:22CODE REVIEWS ARE USEFUL
- 34:37BUT: FLOW DELAY
- 36:04PAIR PROGRAMMING
- 36:21CONTINUOUS INLINE CODE REVIEW
- 38:44CONTINUOUS TESTING
- 41:28NO TEST SUITE DETECTS EVERY ISSUE
- 41:38PIPELINE SPEED = KEY
- 42:40ROLLING UPDATE
- 44:42PIPELINE AS CODE
related talks
44mI Deploy On Fridays
01h 01mI deploy on Fridays (and maybe you should too)
38mFearless Deploys on Friday Evening. Wait; what?!
43mDeploying to Production on Friday
32mHow Your Deployment Strategy Influences How People Code
46mContinuous Deployment Workflows
33mContinuous Delivery Pipelines: How to Build Better Software Faster
31mFireside Chat: Continuous Deployment