Idea Diagnosis
Is this idea worth building, testing or killing? Put it through six questions, in order, before anyone designs anything. Each one needs evidence, not enthusiasm, and the idea ends on one of three verdicts: build, test first, or kill.
Six questions, then a verdict.
- Is the problem real?Someone loses time, money or sleep over it today.
- Who has it?A person you could name, not a market segment.
- What do they do today?The workaround, however ugly.
- Evidence they'd pay?Money or time already spent on the problem.
- Why now?What changed that makes it possible, or urgent.
- Our unfair advantage?What we have that the next team doesn't.
- VerdictBuild, test first, or kill.
What each question needs to pass
The order matters. There's no point asking why now about a problem nobody has, so a failed question stops the run, and the verdict says which one.
Is the problem real?
A real problem costs someone something today: time, money, customers or sleep. A problem you can only describe in the future tense isn't one yet.
Someone describes the cost in their own words, without being led to it.
Who has it?
Narrow enough that you could name ten of them, and find them this week. "Small businesses" fails; "owners of independent clinics who take their own bookings" passes.
A named group you can reach, and the one person inside it who would decide to pay.
What do they do today?
Everyone with a real problem is already doing something about it: a spreadsheet, a group chat, a person paid to chase it. The workaround is your real competitor, and the best evidence the problem is real.
You've seen the workaround, not just heard about it.
Is there evidence they'd pay?
Not whether they say they would. Whether they already spend money or time on the problem, which is what Willingness-to-Pay Evidence separates.
Money or paid time already going into the workaround.
Why now?
Something changed: a model that can now do the hard part, a rule, a shift in how people buy. Without a change, the question is why nobody has done it already.
A specific change you can date, and why it opens the door now.
What's our unfair advantage?
Something the next team can't copy by starting tomorrow: access to the buyers, data nobody else holds, a relationship, a distribution route. Being first and working hard don't count.
An advantage you could describe to a rival without losing it.
Verdict
Build when all six pass. Test first when the idea holds but one answer is a belief rather than evidence, with the cheapest test that would settle it. Kill when the first three fail, or the fifth and sixth both do.
One verdict, in writing, with the question that decided it.
An AI assistant that writes quotes for electricians.
Illustrative example A made-up idea, run the way we'd run a real one. Not client work.
Real, and named
Independent electricians write quotes in the evening, after a day on site, and lose jobs to whoever quotes first. They describe it unprompted. Pass, and pass.
The workaround
A template in a notes app, copied and edited on the phone, and photos sent separately. Seen, not just described. Pass.
Evidence they'd pay
Some pay an office helper to type quotes up. That's money already spent on the problem. Pass.
Why now, and our advantage
Models can now turn site photos and a voice note into a priced quote. But in this illustration nobody on the team has a route to electricians, so the sixth question is a belief. Verdict: test first, by finding out whether a trade supplier would put it in front of its customers.
Where it goes wrong
- Misread
Answering with opinions
Why Every question can be answered with confidence and no evidence. The framework only works when each answer points at something you saw.
- Misread
Starting at "why now"
Why A new model makes many things possible. Possible isn't the same as wanted, which is what the first four questions check.
- Misread
Treating "test first" as a soft yes
Why Test first means one named question is still open. It needs a test with a result that could kill the idea, or it's a build by another name.
In the Journal: Is it worth building? Decide before you pay and Ideate: is the problem real?
Published 28 September 2026. Graylemon original, approved September 2026.