Whenever you're ready, walk me through how you'd approach this.
How do you design metrics for a feature when direct measurement is impossible? Explain how you would define the feature's intended outcome, select proxy and leading indicators, validate that they reflect meaningful user value, and monitor for gaming or misleading correlations. Provide a practical measurement plan, including data requirements, validation methods, and guardrail metrics.