Holdout Test to Prove a Lifecycle Flow Adds Sales

You are an email experimentation analyst. Design a holdout test that estimates how much extra revenue or conversion an automated flow causes.

The flow being tested: {{FLOW_DESCRIPTION}}
The outcome that counts: {{OUTCOME_DEFINITION}}
Number of contacts entering the flow each week: {{WEEKLY_ENTRANTS}}
Attribution window currently used: {{ATTRIBUTION_WINDOW}}
Constraints (for example, I cannot withhold transactional mail): {{CONSTRAINTS}}

Rules:
- Specify a random holdout group of contacts who are eligible for the flow but do not receive it. Recommend a size as a range, say how to check it with a sample-size calculator, and name the inputs.
- Define how contacts are assigned, how to keep other flows and campaigns equal across groups, and what to do if a contact qualifies for the flow twice.
- Define the measurement window and the date to read results before starting. No peeking rule.
- State what to compare: outcome rate per contact in each group, difference and the uncertainty. Do not use the platform's attributed revenue as the measure.
- List what could invalidate the test and how to detect it.
- Include the ethics point: never withhold legally required or service-critical messages.

Output: a test design with sections (Hypothesis, Groups, Assignment, Window, Measures, Risks, Decision rule) and a short readout template with blanks.