Most validation checklists are a set of good questions in no particular order, which means people answer the interesting ones first. The order is worth more than the items. An afternoon settles whether a problem is documented; it takes months to discover you cannot reach the buyer — so the cheap eliminations go first, and most ideas never reach the expensive questions at all.
There are two lists, and they are not the same list
This trips people up, including us: publishing one of these as "the checklist" was wrong in both directions and got caught in review.
If the idea is yours, it goes through 12 filters. If we sourced the candidate, triage applies 10. They ask many of the same questions, and the differences are not cosmetic. The single-idea list adds two the sourced one has no business asking — whether the thing pays for itself inside a year, and whether it could ever be sold or handed off — because those are questions about a business you intend to run. And it does not apply the pain anchor at all: there is no corpus to anchor a pasted idea to, so demanding one would kill every idea for the crime of arriving from outside.
Sourced candidate · 10 checks
- 1Is there a real, documented problem behind it?
- 2Is there a path to money that repeats?
- 3Is there a repeatable way to reach the buyer?
- 4Can it charge enough to be worth one person's time?
- 5Can it run without hand-holding every customer?
- 6Does it survive if one platform changes its rules?
- 7Is there a product left if an AI vendor ships this feature?
- 8Can you, specifically, sell and operate it?
- 9Can it launch without licences or compliance audits?
- 10Can it sell without legal review and a sales team?
An idea you bring · 12 checks
- KF-01Can it make money?
- KF-02Can it charge on repeat?
- KF-03Can it charge enough?
- KF-04Can you reach buyers?
- KF-05Low hands-on support?
- KF-06Safe from one platform's control?
- KF-07A real product, not just an AI wrapper?
- KF-08Can you sell and run it?
- KF-09No heavy compliance blocker?
- KF-10Sells without 'book a call'?
- KF-11Pays for itself within a year?
- KF-12Can it be sold or handed off?
Both lists are imported from the module the gate itself reads, and a test asserts the published list covers exactly the filters the live prompt names — so a check we kill on cannot be one we fail to publish.
Notice what is absent from both. Nothing asks whether the idea is interesting, whether the technology is impressive, or whether the market is large. Not because those do not matter to you, but because none of them is a survival constraint, and mixing preferences into a filter is how a filter stops filtering.
Cost order, and why it is not difficulty order
Sort the questions by what it takes to answer them honestly and three tiers appear.
An afternoon. Is there a documented problem — someone describing it in their own words, in public, unprompted. Can it charge enough. Does it survive one platform changing its rules. These are desk research. You are looking for something that either exists or does not.
A week or two. Is there a repeatable way to reach the buyer. Can it run without hand-holding every customer. These need a real attempt: post something, price something, watch what happens.
Months, or a launch. Whether it pays for itself inside a year. Whether anyone would buy it from you. You cannot desk-research these, which is exactly why they belong last — an idea that fails an afternoon question has cost you an afternoon.
The temptation runs the other way. The expensive questions are the interesting ones, they feel like real work, and answering them produces artefacts you can show people. Cost order is dull and it is the only ordering that saves anything.
What the list actually did to one run
A checklist is a claim about what you would reject. The published run is that claim, settled: 36 of 69 candidates dropped, each with the check that did it.
Every check that killed at least one candidate, most-cited first, counted by the same function the run's own page uses. The top bar is the cheapest question on the list.
The most common cause of death is We found no documented problem — 21 of the 36, more than twice the next reason. That is the first question on the list and the cheapest one to answer, which is the argument for cost order made by the data rather than by us.
The awkward part is also on the chart. Those bars add to 57 reasons against 36 dead ideas, because most of them failed more than one check. The rows do not divide up the drops, and a weak idea is usually weak in several directions at once — which is worth knowing if you were hoping that answering the one objection you have heard would rescue yours.
What a checklist cannot do
It cannot tell you an idea is good. Every check here is a way of failing, and passing all of them means only that no known deal-breaker was found — which is a licence to run a real test, not a verdict. The list is a filter, and a filter's job is to be cheap and to be wrong in the safe direction.
It also cannot be applied honestly to your own idea in your head. The failure mode is not ignorance of the questions; it is answering "is there a documented problem" from memory of a conversation you had, and moving on. Write the answer down with the link in it, or you have not answered it.
The short version
- Two lists: 10 for a candidate we sourced, 12 for an idea you bring. They differ on purpose and neither is "the" checklist.
- Answer them in cost order — an afternoon, then a fortnight, then months. Most ideas die in the first tier and never cost you the rest.
- On the published run the cheapest question killed the most: 21 of 36 drops.
- 57 reasons against 36 dead ideas: weak ideas fail several checks at once, so fixing the objection you have heard usually does not save one.
- Passing everything is permission to test, not a verdict.