# Email Subject Line Test

Compare truthful subject lines for one defined audience using a documented hypothesis, controlled variants, suitable metric, stopping rule, and decision record.

## Usage note

Use this worksheet for one decision with a defined audience and send context. Every variant must describe the email truthfully and preserve material conditions. Follow consent, suppression, privacy, deliverability, and applicable messaging rules. A higher open rate does not justify deceptive urgency, hidden sponsorship, or a subject that the message cannot fulfill.

## How to use this template

1. Define the recipient situation, email task, research evidence, and one wording decision the test will inform.
2. Write truthful variants that differ on the chosen factor while keeping material conditions visible.
3. Pre-register audience assignment, sample approach, primary metric, guardrails, exclusions, duration, and stopping rule.
4. Quality-check the variants and message, then launch only after data, deliverability, and approval owners confirm readiness.
5. Analyze the complete result, document uncertainty and harms, and record the scoped decision for future sends.

## Blank template

### Experiment foundation

- **Experiment ID:** [Durable identifier]
- **Email and lifecycle moment:** [Message and trigger]
- **Recipient audience:** [Eligibility and exclusions]
- **Recipient task:** [What should they understand or do?]
- **Evidence or observation:** [Why test this?]
- **Decision to inform:** [Specific future choice]
- **Hypothesis:** [Audience + change + expected behavior + reason]
- **Experiment owner:** [Name or role]

### Variants and controls

- **Control subject:** [Exact text]
- **Variant subject:** [Exact text]
- **Single difference:** [Specificity, action, framing, or another]
- **Preheader:** [Stable text or documented variable]
- **Sender identity:** [Stable sender]
- **Message body version:** [Exact approved version]
- **Truth and condition check:** [How each subject is fulfilled]
- [ ] No variant uses false urgency, personalization, or implied relationship.
- [ ] Accessible characters and meaningful words appear early.
- [ ] Audience, timing, sender, and body are controlled.

### Measurement plan

- **Assignment method:** [Randomization and split]
- **Required sample or run:** [Method and owner]
- **Primary metric:** [Metric tied to the decision]
- **Guardrail metrics:** [Delivery, complaints, unsubscribe, task failure, or another]
- **Exclusion rules:** [Bots, duplicates, internal traffic, or another]
- **Stopping rule:** [Written before launch]
- **Segment analysis:** [Only pre-planned relevant groups]
- **Data owner:** [Analyst or team]

### Result and decision

- **Run dates:** [Start and end]
- **Variant counts:** [Delivered population]
- **Primary result:** [Value and uncertainty]
- **Guardrail result:** [Any harm or anomaly]
- **Operational issue:** [Tracking or delivery concern]
- **Decision:** [Adopt / Retain control / Follow up / Inconclusive]
- **Scope of learning:** [Where it may apply]
- **Next action and owner:** [Specific change]