Metric design patterns for AI features

This is Lesson 4.2: Metric Design Patterns. The previous lesson covered classifying metrics into two buckets. Blocking metrics gate ship decisions. Optimization metrics are tracked over time. That was the concept. This lesson makes it operational. The shift: instead of inventing a new metric every time a new failure mode appears, the approach is to apply measurement archetypes. These are reusable templates. Each one specifies what to measure, at what granularity, and which metrics should block a release versus which ones to track. Why does this matter? Teams with perfectly good metrics can still be unable to answer "can we ship?" They have the data but lack a system. That is what this lesson builds: the system that turns individual metrics into a coherent measurement framework. By the end, the deliverable is a portfolio-ready artifact that makes ship decisions obvious.