Decide the sample size before you run the test

I want to test {{change}} on {{surface}}. Current baseline: {{metric and its current rate or mean, plus its variance if you have it}}. Traffic: {{visitors or events per day eligible for the test}}.

Work out, showing your reasoning in plain language:
1. The smallest effect that would actually change our decision — ask me for it if I haven't said, and explain why 'any improvement' is not an answer.
2. The sample size per arm for that effect at 80% and 90% power, and how many days each implies at my traffic.
3. Whether this test is feasible at all. If it needs more days than {{our patience}}, say so and give me the options: a bigger change, a proxy metric closer to the surface, or not testing.
4. The guardrail metrics that must not degrade, and their thresholds.
5. The exact stopping rule, plus what I should do when someone asks to peek on day three.

End with the one-paragraph test plan I can paste into the ticket.