How an experiment gets designed and run
Every step of a Heatseeker experiment, from choosing one of eight templates in your workspace to reading a statistically significant result, and what you are asked to approve along the way.
Every Heatseeker experiment starts from a template and ends with a decision you can defend. Here's what happens at each step, and what you're asked to approve.
1. Choose a template
Start a new experiment in your workspace and pick one of eight templates:
- Value Proposition Test: which promise actually makes people act
- Messaging Test: which words land, and in which market
- Pain-Point Test: which problem they'll pay to have fixed
- Concept Test: which idea deserves the budget
- Feature Test: which capability tips the decision
- Full Creative Test: which creative earns attention
- Buying Driver Test: what really closes the sale
- Product Demand Test: whether anyone wants it, before you build it
Each carries its own methodology, so the audience logic, variant structure and success measure are already set up for that kind of decision.
2. Set the goal
Tell Heatseeker the decision you're trying to make and who you're trying to reach. This is the only part that needs your knowledge of the business.
3. Heatseeker designs the experiment
The AI selects the audience, writes the variants and assembles the experiment. You'll see a draft in your workspace with everything laid out.
4. Approve the creative
Review the variants, edit anything you want changed, and approve. Nothing goes to market without your sign-off.
5. Choose live or synthetic
Run it live with real customers for behavioral evidence, or against synthetic personas trained on your own data for a read in minutes. Most teams use synthetic to narrow the field, then go live to confirm.
Live experiments run on Meta and LinkedIn, so those accounts need to be connected first. See Connect Meta to Heatseeker and Connect LinkedIn to Heatseeker.
6. The launch checklist
Heatseeker won't launch an experiment that isn't ready. Outstanding requirements are listed on the draft, and the launch stays blocked until each one is met.
7. Read the result
Results come back with the winning variant, the margin, and whether the difference is statistically significant. Every claim is drillable to the underlying responses.
You can also ask about a finished experiment from your own assistant. See Connect to Heatseeker with MCP.