Design an A/B test for a product feature

**Role:** You are a Product Manager designing an A/B test for {feature_being_tested} on {product_name}.

**Context:**
Poorly designed A/B tests lead to false conclusions that drive the product in the wrong direction. A rigorous experiment starts with a clear hypothesis, defines success metrics before running the test, calculates the required sample size to achieve statistical significance, and establishes guardrails to protect against harm during the test.

**Task:**
Design a complete A/B test plan. Apply step-by-step reasoning: (1) define the hypothesis, (2) determine the primary and guardrail metrics, (3) calculate required sample size, (4) define the control and variant, (5) establish stopping rules.

**Input Available:**
- {feature_being_tested}: What is being tested
- {product_name}: Product name
- {hypothesis}: Your belief about what will happen and why
- {primary_metric}: The one metric that determines success or failure
- {current_baseline}: Current value of the primary metric
- {minimum_detectable_effect}: The smallest improvement that would be worth shipping
- {daily_users}: Number of users who encounter this surface daily
- {guardrail_metrics}: Metrics that must not degrade during the test

**Output Format:**
1. Hypothesis statement: If [we do X], then [Y metric] will [increase/decrease] by [Z%] because [mechanism]
2. Test design: Control (what users see today) vs. Variant (what you're testing)
3. Primary success metric with definition and measurement method
4. Guardrail metrics with acceptable variance thresholds
5. Sample size calculation: Required users per variant to achieve 95% confidence
6. Estimated test duration: Based on daily users and required sample size
7. Audience allocation: What % traffic to each variant and why
8. Stopping rules: Pre-defined conditions to stop early (positive and negative)
9. Analysis plan: How you will analyze results and interpret them
10. Rollout plan: What happens if the variant wins / loses / is inconclusive

**Guardrails & Quality Control:**
- Do not peek at results before reaching required sample size — this inflates false positive rates
- Guardrail metrics must be defined before the test starts, not after analyzing results
- Run the test for a minimum of 1 full business cycle (usually 2 weeks) to avoid day-of-week bias
- An inconclusive result is valuable information — build a clear interpretation for this case