Before I published Feed a Starving Crowd, I spent about six months arguing with myself over the title. Then I spent $60 on Facebook ads and had an answer in 48 hours.
That's the lesson I want you to take into this tool. Guessing is slow. Testing is fast. And the best test always involves real people spending real attention or real money.
So why build a synthetic panel at all?
Because most people skip testing entirely. A simulated panel costs nothing and takes a minute. It won't tell you what customers will pay, but it will often surface the obvious objection you missed, the confusing line in your offer, or the price that looks wrong next to the others. It's a dress rehearsal. And then it hands you the script for the real performance.
How do you use the Synthetic Customer Panel?
- Describe your offer in two or three sentences, the way you'd explain it to a customer.
- Name your audience as specifically as you can. "Office workers aged 30 to 55 with recurring back pain" beats "adults."
- List two to four price options, with the period ("$89, $129 or $179 a month").
- Choose three to eight personas and run the panel.
You get a table of simulated personas (reaction, the most they'd pay, their main objection and how likely they are to buy), a synthetic Sean Ellis estimate and a verdict. Everything is labeled as simulated. Underneath, the tool builds a real-world test kit for your offer: five interview questions, four price survey questions and one commitment test.
What you type is sent to an AI model to simulate the panel and isn't stored by this site.
The numbers behind it
The synthetic Sean Ellis estimate borrows the product/market fit test from the book. With real customers you ask how they'd feel if they could no longer use the product.
Sean Ellis score = "Very disappointed" answers ÷ All answers × 100 (40% or more suggests product/market fit) Price spread = Highest price option ÷ Lowest price option
A simulated score is a talking point, not a measurement. The price spread is more useful than it looks: if your top option is less than about 1.4 times your lowest, you'll struggle to learn where interest drops off.
A worked example: a physio membership
The preloaded offer is a physio membership with two sessions a month, a weekly pilates class and an exercise app, for CBD office workers, at $89, $129 or $179 a month.
The price ladder check notes a 2.0× spread, a sensible range, and flags $129 as the likely "most popular" anchor. A typical simulated panel will raise objections like "my health fund already covers some physio" or "two sessions won't be enough in a flare-up." Those are exactly the things to probe in real interviews, and possibly to answer in the offer itself.
How should you read a simulated panel?
Treat every number as directional. Simulated personas are generated by an AI model from patterns in its training data. They don't have budgets, bad days, partners who veto purchases or a competitor's offer sitting in their inbox. They also tend to be more polite and more consistent than real people.
Use the panel for three things only: to find objections worth testing, to tighten your wording, and to choose which prices to put in front of real customers. Never use it to set your final price or to forecast sales.
Then do the real thing. Twenty conversations, a short price survey and a pre-sale will teach you more than any simulation.
What to do next
- Run real interviews with the willingness-to-pay interview script.
- Send the willingness-to-pay survey and score it with the Van Westendorp calculator.
- Read how to measure willingness to pay and the Van Westendorp method.
- Once you have customers, run the real Sean Ellis 40% test.
Frequently asked questions
Are synthetic personas real market research?
No. They're a rehearsal. The personas are simulated by an AI model, so they reflect patterns in its training data rather than the opinions of your actual customers. They're useful for spotting obvious objections and testing your wording before you spend money, but they can't tell you what real people will pay. Real interviews, surveys and sales always win.
How many personas should I simulate?
Six is a good default. Three is enough to sense-check wording; eight gives more variety of objections. More personas don't make the result more accurate, because they all come from the same model. If you want a bigger sample, change the audience description or the offer and run the panel again, then compare what changes.
Why does the panel give a 'synthetic Sean Ellis' score?
It's a simulated version of the product/market fit question from the book: how would you feel if you could no longer use this? With real users, 40% or more answering 'very disappointed' suggests product/market fit. A simulated score is only a talking point. Run the real survey with real users before you draw any conclusion.
What should I do after running the panel?
Take the objections it raised and test them with real people. Use the five interview questions the tool writes for your offer, send the four Van Westendorp price questions to a short list of prospects, and ask for a real commitment such as a deposit, pre-sale or waitlist sign-up at your middle price. Money from real people is the only proof.
Built on The Subscription Playbook
This tool turns frameworks from Robert Coorey's book into something you can use in minutes. Robert is the author of The Subscription Playbook and a Director at BPAG Accountants. More about Robert.