Mary and Tom Poppendieck's lean software principles adapted for cloud

• 1. Eliminate waste — Map your deployment pipeline as a value stream. Every step that is not delivering value to users is a candidate for elimination or automation. Ask: "What would happen if we removed this step?" If the answer is "nothing bad," remove it.
• 2. Amplify learning — Make feedback loops as short as possible. Automated tests that run in 2 minutes, not 45. Feature flags that enable A/B testing in production instead of months-long experiments. Blameless post-mortems that generate improvements, not blame.
• 3. Decide as late as possible — Keep architecture options open. Avoid premature optimization. Do not choose a database engine before you understand your access patterns. Do not build caching before you have measured where the latency is.
• 4. Deliver as fast as possible — Small, frequent deploys reduce risk (smaller blast radius), improve feedback (know what caused the bug), and increase velocity. The goal: multiple production deploys per day, not per quarter.
• 5. Empower the team — Decisions are best made by those with the most context — the engineers building the system. Avoid centralized gatekeepers (single ops team managing all deployments, all infrastructure, all access). Platform engineering solves this: self-service infrastructure with guardrails.
• 6. Build integrity in — Quality cannot be inspected in after the fact — it must be built in at every step. Automated tests, linters, security scans, infrastructure-as-code — not a QA team at the end of the pipeline.
• 7. See the whole — Optimize the entire system, not individual components. A faster CI pipeline that creates a slower deployment process is not an improvement. Measure DORA metrics: deployment frequency, lead time for changes, change failure rate, time to restore service.