Decision-making under uncertainty

Week 6 is the decision-making week. The previous five weeks covered building evaluation infrastructure: instrumentation, metrics, judges, oracles, experiments, and launch readiness. The result is a full pipeline that reveals whether a change made things better, worse, or had no impact. But having all the data does not make the decision easy. It often makes it harder, because now the tradeoffs are visible. Quality improved on two dimensions but regressed on a third. Some user segments benefit while others do not. Blocking metrics pass but optimization metrics are mixed. And someone in the room has to say "ship it" or "hold." That is what this week addresses. Not more measurement. Decision-making. This lesson introduces the Ship Decision Framework, a structured approach for converting conflicting signals into justified decisions. The key word is "justified." Not perfect. Justified, meaning the reasoning explains why this path was chosen, what evidence supports it, what risks were accepted, and what happens if things go wrong. Most teams fail here not because they lack data but because they lack a decision vocabulary. They have two words, ship and don't ship, and that binary forces every nuanced situation into a false choice. By the end of this lesson, the framework provides six decision types instead of two, a rubric for matching evidence to decision type, and a tradeoff transparency template that makes reasoning auditable.