Evaluate AI-assisted ads creation feature

You’re launching an **AI-assisted ad creation** tool for advertisers (it suggests copy/creative and helps generate new ads). You need to evaluate whether it is beneficial and safe to ship broadly. ### Prompt 1) What are the **primary success metric(s)**, **diagnostic metrics**, and **guardrail metrics** you would use? Explain tradeoffs and how you would avoid “winning” on a proxy while harming the ecosystem. 2) How would you **design an experiment** to measure **incremental impact** of the tool? Be specific about: - unit of randomization (advertiser, campaign, ad account, geo, auction bucket, etc.) - opt-in/partial adoption and how to handle it - duration, power/MDE, and variance reduction ideas - risks like interference/marketplace effects and novelty effects 3) If the tool increases total spend but decreases downstream user experience (e.g., more complaints), how would you decide whether to launch?