For solo and small-team builders · research to 5 October 2026
Before you build it, find out if someone will actually pay for your AI idea.
Here's how it works. You answer six short questions about your own idea, each taking about ten minutes. At the end you get one word back: TEST (try selling it), REPOSITION (change who it's for or what it does) or KILL (stop before you spend more time and money on it).
You have probably seen small AI tools claiming US$15k–160k a month. Those are sales figures the founders reported themselves - not income, and not audited. The ones that last usually do one dull job very well, and that, then, is what I'll look for in your idea.
What will you have at the end?
Think of each step as one sentence you finish about your idea, plus a one-word result. By step 6, then, you have a one-page card you can copy, print or send to someone you trust.
| Step | The blank you finish |
|---|---|
| 1 | My buyer's best free alternative is ___ |
| 2 | After costs, I need ___ left per month |
| 3 | If the model were free, they would still pay me for ___ |
| 4 | I would hesitate to leave because ___ (in the customer's voice) |
| 5 | Six questions in one paragraph, no score → TEST, REPOSITION or KILL |
| 6 | The first paid offer I will make, before I build another feature, is ___ |
Is this for you?
Yes, if
You build alone or with one or two others. You have an AI idea, maybe half-built, and you want to know whether someone will pay for it - and keep paying.
Not if
You want market sizing, investor slides or a success rate. The research doesn't have a success rate, and this guide won't invent one.
What happens to what you type?
It stays in this browser on this device. The next step reads it back so you don't retype anything. Nothing is sent to me or anyone else.
"I don't know yet" is a fair answer on every step. It becomes a small test for you to run, not a failing grade.
Questions you might have
Is my idea too small to test?
No. In fact, smaller is easier. If one person would pay you £49 to solve one problem for them, that is enough to test. You don't need a big market to start, just one buyer who says yes.
What happens to what I type here?
It is saved in this browser on this device (local storage), so each step can show your earlier answers. Nothing is sent to Anvil Analytics or anyone else. "Clear my saved answers" at the bottom of every page deletes it.
Is this free?
Yes. All six steps and your field card are free, with no sign-up. The only paid thing is the optional clinic on step 6, and booking for it isn't open yet.
Do I need to be technical?
No. You answer in plain words about your buyer and your money, the way you'd explain it to a friend. There is no code and no AI jargon to learn first. Where a term matters (for example "floor"), the page explains it as you go.
How long does it take?
About ten minutes per step, so roughly an hour in total. You don't have to do it in one sitting. Stop whenever you like and come back later on the same device, and your answers will still be here.
What if I get KILL?
You stop, and you have saved yourself months of building something nobody pays for. KILL tells you which answer caused it. You can change that part of the idea and run the steps again, or drop it and test your next idea.
Can I skip straight to the end?
Yes, but step 5 and step 6 work better with your earlier answers. If you skip, step 5 can only say NOT ENOUGH EVIDENCE until you mark at least two questions as Supported.
What is the clinic, and what does it cost?
An optional paid review of one idea by a real person: 30–45 minutes, written or on a video call, ending in a written TEST, REPOSITION or KILL with reasons. The price is £49, one payment. Booking isn't open yet - the button on step 6 says so.
Can you trust the numbers here?
Only as far as the label says. Every money figure carries a label showing how well it is proven, because most public "AI wins" are founder claims nobody checked.
Founder-reported the founder said it, nobody audited it · Secondary someone else reported it · Verified announcement dated primary source · Editorial target a goal I chose, not a market fact · Unknown the evidence doesn't say
You now have:
- A map of the six blanks and what each step gives you.
- Permission to answer "I don't know yet".