Outreach A/B Test Plan
Test messages without fooling yourself.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Design an outreach experiment. Channel and current performance: [CHANNEL_AND_BASELINE]. Audience size available: [AUDIENCE_SIZE]. Variables I want to test (subject, opening line, call to action, length, offer): [VARIABLES]. Constraints (compliance, tools): [CONSTRAINTS]. Return valid JSON only with keys: "goal", "primary_metric" (for example positive reply rate), "secondary_metrics", "hypotheses" (array of objects: "id", "statement", "variable", "expected_direction", "reason"), "design" (object: "variants", "split_method" such as random assignment, "sample_size_per_variant" with a note that it is approximate, "duration_days", "send_times_controlled": true or false, "audience_overlap_control"), "guardrails" (array: stop if complaint rate rises, honest messaging only, opt-out honored), "analysis_plan" (how to compare, minimum difference worth acting on, caution about multiple tests), "decision_rules" (array), "risks" (array), and "learning_log_fields" (array). After the JSON, in plain text: explain in simple terms why small samples can mislead (give an example with numbers), how many replies are needed before drawing conclusions (state it as a rough rule of thumb), why to test one variable at a time, and how to record learnings. Do not promise outcomes, and keep tests ethical (no deceptive variants).
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.