Session abstract
What you’ll learn
Modern CI/CD systems assume stable inputs: clear requirements, predictable sequencing, and decoupled deployments. But in real-world e-commerce systems, engineering teams often operate under unstable conditions — late-breaking requirements, cross-team dependencies, and content tightly coupled to release pipelines. In this talk, an analysis demonstration of real delivery data will show how defect rates shift dramatically under different intake conditions — even with the same team, codebase, and tooling. In one case, a system operating under stable conditions achieved a 5:1 story-to-bug ratio. Under unstable intake conditions, that ratio degraded toward 1:1. We’ll explore the underlying system constraints driving this behavior from real-world experience and then walk through concrete engineering solutions, including: - Treating Definition of Ready (DoR) as a validation layer for work entering the system - Decoupling content from code using CMS-driven architectures - Isolating releases with feature flags and environment strategies - Designing for dependency-aware delivery across distributed systems - Using delivery metrics (e.g., story-to-bug ratio) as observability signals for system health This talk is for engineers and tech leads building systems that must perform reliably under real-world pressure — where inputs are often incomplete, late, or constantly changing.