Measurement

Shipping is not the outcome; the behavior change is. This skill closes the loop: it takes the success metric written in the frame, compares against the baseline, and turns the answer into evidence for the next cycle. Without it the team is a factory that never learns whether its products work.

The contract was written in the frame

The frame declared a target behavior, a success metric, and a baseline. Measurement holds the work to that contract:

• No baseline, no verdict. If the baseline wasn't captured before ship, the first job is instrumenting one now and the honest report is "unmeasurable this cycle".
• Measure the frame's metric, not a flattering neighbor. "Engagement is up" doesn't answer "did trial users reach their first deploy faster?"
• Wait for the novelty decay. Week-one numbers measure curiosity. Compare stable windows: same weekdays, enough volume, seasonality acknowledged.

Reading the result

• Moved as predicted: the frame's assumption graduates to fact in the evidence ledger, with the numbers as source.
• Didn't move: the feature works and the hypothesis was wrong. That's information, not failure; the assumption it kills goes back in the ledger marked falsified. The expensive error is quietly keeping the belief.
• Moved the wrong way: check the guardrail metrics you should have paired with the success metric (support tickets, task time, error rate, churn on the adjacent flow). Every success metric needs at least one guardrail, or you'll optimize it at the product's expense.
• Can't tell: volume too low or instrumentation missing. Say so. A shrug with numbers attached beats a story without them.

Segment before concluding: an average that's flat is often one audience up and another down, and the frame named which audience mattered.

Qualitative closes the why

Numbers say what changed; they never say why. Pair the metric read with the cheap qualitative sources: support tickets mentioning the feature, session recordings of the changed flow, the in-product search terms. Route what you learn to the owning skill: wrong words to content-design, wrong flow to interaction-design, wrong bet to the next frame.

The loop

The report is one page: metric, baseline, result, verdict, what enters the evidence ledger, and the single next decision (iterate, expand, revert, or leave it alone). File the ledger entries; the next product-evidence pass starts from them. That's the difference between a design team and a design factory: the factory ships; the team finds out.