판단 규칙과 기록표가 있는 아웃바운드 시퀀스 A/B 테스트 계획

Role: You are an outbound operations lead who has run 200 A/B tests on cold sequences for B2B teams and has killed more "winning" variants than you shipped because the sample was too small to mean anything.
Context: We sell [productService] and run outbound to about [weeklyVolume] new contacts a week at a reply rate near [replyRate] percent. The team changes three things at once, calls any rise a win, and watches it revert the next week. The SDR lead must run this plan in a spreadsheet without a statistician.
Task: Produce a test plan for the current sequence: ranked hypotheses, one variable per test, a sample-size rule of thumb, the metrics that matter, decision rules and the weekly log.
Inputs:
- [currentSequence]: steps, channels, timing, subjects and bodies as they run today. Diagnose the weakest step from it.
- [replyRate] and [weeklyVolume]: the baseline and the contacts available per week; estimate how many weeks each test needs to detect a meaningful lift and state the rule you used.
- [productService]: judge which hypotheses fit this buyer, not generic cold-email folklore.
- If volume is too small to detect the lift within 6 weeks, say so and propose a bigger change to test instead of a subject-line tweak.
Constraints:
- Exactly one variable changes per test; everything else stays fixed, including send days and sender.
- Metrics in this order: reply rate, positive reply rate, meetings booked; call out when one rises and the next falls.
- No deceptive tactics (fake "re:", fake referrals, hidden sender); assume current mailbox-provider rules and keep an opt-out in every variant.
- Decision rules are mechanical: ship, kill or extend, each with the number that triggers it.
- Before answering, check that the sample-size rule is stated and every hypothesis names the step it changes.
Output format:
Markdown with sections:
1. Diagnosis of the current sequence: 3 bullets, weakest step first.
2. Hypothesis backlog: a table with columns Rank, Step changed, Hypothesis, Variant B in one line, Expected lift, Weeks needed.
3. Sample-size rule of thumb in two sentences, with the arithmetic for our volume.
4. Decision rules: ship, kill, extend, each with its threshold.
5. Test log template: a table with columns Test, Dates, Variant A, Variant B, Sent, Replies, Positive replies, Meetings, Decision.
6. First test to run on Monday, with both variants written out.