AI Marketing Experiment Workflow

5 steps to a reviewable draft

1. 01

   Name the decision at risk

   Write the choice the test should inform. If no decision will change after the result, do not run an experiment.

   Result: One decision statement with an owner.

2. 02

   Define the test boundary

   Record the audience, channel, offer, time window, fixed conditions and operational limits. Call out what cannot stay controlled.

   Result: A practical test boundary rather than a vague learning goal.

3. 03

   Write the experiment card

   Ask for a plan that makes the hypothesis, measurement and decision rule inspectable before the work begins.

   Result: A test card that makes a future result easier to interpret.

   Your prompt

   Copy prompt

   Create a marketing experiment card from the supplied inputs.
   
   Decision: {{decision}}
   Audience and channel: {{audience_channel}}
   Proposed change: {{change}}
   Current evidence: {{evidence}}
   Measurement limits: {{measurement}}
   
   Include: hypothesis, audience, change, fixed conditions, primary metric, guardrail metrics, duration, required sample or exposure condition, risks, stop rule, result interpretation and next decision. Mark unknown thresholds as [SET BEFORE RUN]. Do not promise statistical significance, causality or commercial impact when the inputs do not support it.

   Example input

   Decide whether a workshop invitation should lead with reporting time saved or client communication. Email audience is existing users; the send window is ten days.

   Example output

   Hypothesis: [TO TEST] reporting-time framing earns more registrations. Guardrail: unsubscribe rate. Stop rule: pause if complaints exceed the agreed threshold.

4. 04

   Check the measurement plan

   Ask a channel or data owner to verify that the primary metric, guardrails and exposure condition can actually be observed.

   Result: A plan with feasible measurement and visible risks.

5. 05

   Commit the run sheet

   Assign owners, record the start condition and save the test card with the campaign or research record.

   Result: An experiment that can be reviewed after it runs without reconstructing the original plan.