DEVOPS DAYS MTL 2018 - Lenore Adam
summary
Describes turning CD pipelines into a measurement engine by using feature flags to separate deploy from release, moving from whitelists to randomized exposure (example: ramping to 20–50% of users) and keeping a global kill switch to avoid emergency patches. Gives concrete tools and practices: targeted rollouts, a randomization engine, A/B tests with t‑tests for statistical significance, AAA sanity checks, and real case studies (manual backend for a tiny cohort and a freemium trial tweak) that prove whether changes justify full investment.
AI-written summary · about
chapters (14)
- 0:00Controlled rollouts to experiments
- 2:29Feature flags as deployment
- 4:56Phased rollouts with targeting
- 7:50Kill switches for quality
- 9:13Testing in production for QA
- 11:42Foundation for experiments
- 13:09Experimentation as innovation
- 15:57Redefining the Minimum Viable Product
- 16:37Develop locally analyze globally
- 18:03Empowering teams with data
- 20:10Hypothesis-driven development
- 23:40Statistical significance
- 24:42Path to trust and acceptance
- 26:49Closing and platform overview
related talks
53mDeliver Results, Not Just Releases: Control & Observability in CD
38mProgressive Delivery: Patterns & Benefits of Decoupling Deploy from Release
06mDeploying Radically, Democratizing Release
30mDevOps Days MTL 2019 - Martin Deslongchamps
19mThink Prod Up Front: Shifting Left on Production Excellence
02mAdvanced Deployment Pipeline Techniques.
12mRevolutionizing Software Development
06mDeploying rapidly for continuous integration